Skip to content
Trazum

Trazum

Models and prices

Every model the catalogue prices, grouped by provider, with the rates the product bills with and the date each provider’s page was last read.

Rates are US dollars per million tokens, as the catalogue holds them. Nothing on this page is computed: a cache or batch price is shown as the multiple the catalogue states, not as a figure.

Oldest provider review: 2026-11-05

Showing 91 of 91 models

Anthropic22 models

Prices last read 2026-11-16·platform.claude.com/docs/en/about-claude/pricing

ModelInputper 1M tokensOutputper 1M tokensContextCache readBatch
Claude Fable 5.2claude-fable-5-2$8.00$40.001,000,0000.025x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Bands
  • Fast mode (usage.speed “fast”)·$16.00 / $80.00fast
Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Cache hits cost 2.5% of the input price ($0.20). Fast mode (usage.speed "fast") bills $16/$80. Requires 30-day data retention.
Claude Mythos 5.2claude-mythos-5-2Not offered as an alternative$8.00$40.001,000,0000.025x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Bands
  • Fast mode (usage.speed “fast”)·$16.00 / $80.00fast
Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Available only to organizations verified through Anthropic's verification programs. Fast mode (usage.speed "fast") bills $16/$80. Requires 30-day data retention.
Claude Fable 5.1claude-fable-5-1$10.00$50.001,000,0000.025x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Legacy: superseded by Claude Fable 5.2, which costs less ($8/$40). Cache hits cost 2.5% of the input price ($0.25). Requires 30-day data retention.
Claude Mythos 5.1claude-mythos-5-1Not offered as an alternative$10.00$50.001,000,0000.025x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Available only to organizations verified through Anthropic's verification programs. Requires 30-day data retention.
Claude Fable 5claude-fable-5$10.00$50.001,000,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Legacy: superseded by Claude Fable 5.1 at the same prices. Requires 30-day data retention.
Claude Mythos 5claude-mythos-5Not offered as an alternative$10.00$50.001,000,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Available only to organizations verified through Anthropic's verification programs. Requires 30-day data retention.
Claude Opus 5.5claude-opus-5-5$4.00$20.001,000,0000.05x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Bands
  • Fast mode (usage.speed “fast”)·$8.00 / $40.00fast
Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Cache hits cost 5% of the input price ($0.20), against 10% on Opus 5 and earlier. Fast mode (usage.speed "fast") bills $8/$40.
Claude Opus 5claude-opus-5$5.00$25.001,000,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Bands
  • Fast mode (usage.speed “fast”)·$10.00 / $50.00fast
Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
512-token cache minimum: caches prompts that would miss on Opus 4.6. Fast mode bills $10/$50.
Claude Opus 4.8claude-opus-4-8$5.00$25.001,000,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 1,024 tokens

write 1.25x for 5 min

write 2x for 1 h

Bands
  • Fast mode (usage.speed “fast”)·$10.00 / $50.00fast
Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Fast mode bills $10/$50 until it is removed from this model on 2026-11-30.
Claude Opus 4.7claude-opus-4-7$5.00$25.001,000,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 2,048 tokens

write 1.25x for 5 min

write 2x for 1 h

Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Claude Opus 4.6claude-opus-4-6$5.00$25.001,000,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 4,096 tokens

write 1.25x for 5 min

write 2x for 1 h

Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Claude Opus 4.5claude-opus-4-5$5.00$25.00200,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 4,096 tokens

write 1.25x for 5 min

write 2x for 1 h

Web search
$10.00 per 1,000 searches
Also billed as
  • claude-opus-4-5-20251101
Catalogue note
200K context window and a 64K output cap.
Claude Opus 4.1claude-opus-4-1-20250805Not offered as an alternative$15.00$75.00200,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 1,024 tokens

write 1.25x for 5 min

write 2x for 1 h

Web search
$10.00 per 1,000 searches
Catalogue note
Retired on the Claude API on 2026-08-05; still served on Google Cloud (Google managed) and Bedrock. Anthropic names Claude Opus 4.8 as the replacement.
Claude Opus 4claude-opus-4-20250514Not offered as an alternative$15.00$75.00200,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 1,024 tokens

write 1.25x for 5 min

write 2x for 1 h

Web search
$10.00 per 1,000 searches
Catalogue note
Retired on the Claude API on 2026-06-15; still served on Google Cloud (Google managed). Anthropic names Claude Opus 4.8 as the replacement.
Claude Sonnet 5.5claude-sonnet-5-5$2.00$10.001,000,0000.05x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Cache hits cost 5% of the input price ($0.10), against 10% ($0.20) on Sonnet 5.
Claude Sonnet 5claude-sonnet-5$2.00$10.001,000,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 1,024 tokens

write 1.25x for 5 min

write 2x for 1 h

Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Launched with 2/10 as introductory pricing through 2026-08-31; the scheduled increase to 3/15 on 2026-09-01 was cancelled and 2/10 is the standard price.
Claude Sonnet 4.6claude-sonnet-4-6$3.00$15.001,000,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 1,024 tokens

write 1.25x for 5 min

write 2x for 1 h

Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Claude Sonnet 4.5claude-sonnet-4-5Not offered as an alternative$3.00$15.00200,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 1,024 tokens

write 1.25x for 5 min

write 2x for 1 h

Web search
$10.00 per 1,000 searches
Also billed as
  • claude-sonnet-4-5-20250929
Catalogue note
Deprecated; retires 2026-11-30. Recommended replacement: Claude Sonnet 5.5.
Claude Sonnet 4claude-sonnet-4-20250514Not offered as an alternative$3.00$15.00200,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 1,024 tokens

write 1.25x for 5 min

write 2x for 1 h

Web search
$10.00 per 1,000 searches
Catalogue note
Retired on the Claude API on 2026-06-15; still served on Google Cloud (Google managed) and Bedrock. Anthropic names Claude Sonnet 5.5 as the replacement.
Claude Haiku 5.5claude-haiku-5-5$0.1000$0.50001,000,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 512 tokens

write 1.25x for 5 min

write 2x for 1 h

Bands
  • Prompt above 100,000 input tokens·$0.5000 / $2.50over-100k
Inference geography

us 1.1x

Web search
$10.00 per 1,000 searches
Catalogue note
Priced per request by prompt length: $0.10/$0.50 up to 100,000 input tokens, cache reads and writes included, and $0.50/$2.50 over it.
Claude Haiku 4.5claude-haiku-4-5$1.00$5.00200,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 4,096 tokens

write 1.25x for 5 min

write 2x for 1 h

Web search
$10.00 per 1,000 searches
Also billed as
  • claude-haiku-4-5-20251001
Catalogue note
200K context window and a 64K output cap, smaller than the rest.
Claude Haiku 3.5claude-3-5-haiku-20241022Not offered as an alternative$0.8000$4.00200,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: marked by the caller

from 2,048 tokens

write 1.25x for 5 min

write 2x for 1 h

Web search
$10.00 per 1,000 searches
Catalogue note
Retired on the Claude API on 2026-02-19; still served on Google Cloud (Google managed). Anthropic names Claude Haiku 4.5 as the replacement.

OpenAI36 models

Prices last read 2026-11-16·developers.openai.com/api/docs/pricing

ModelInputper 1M tokensOutputper 1M tokensContextCache readBatch
GPT-6 Astragpt-6-astra$10.00$50.001,050,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1.25x for 30 min

write 2.5x for 12 h

Bands
  • Prompt above 272,000 input tokens·$20.00 / $75.00over-272k
Service tiers
  • flex·$5.00 / $25.00
    • Prompt above 272,000 input tokens·$10.00 / $37.50over-272k
  • fast·$20.00 / $100.00
    • Prompt above 272,000 input tokens·$40.00 / $150.00over-272k
  • ultrafast·$60.00 / $300.00
    • Prompt above 272,000 input tokens·$120.00 / $450.00over-272k
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and cache and 1.5x output for the whole request. Cache writes cost 1.25x input for the default 30-minute entry and 2.5x for a 12-hour one (Responses API).
GPT-6.1 Solgpt-6.1-sol$2.00$10.001,050,0000.05x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1.25x for 30 min

write 2.5x for 12 h

Bands
  • Prompt above 272,000 input tokens·$4.00 / $15.00over-272k
Service tiers
  • flex·$1.00 / $5.00
    • Prompt above 272,000 input tokens·$2.00 / $7.50over-272k
  • fast·$4.00 / $20.00
    • Prompt above 272,000 input tokens·$8.00 / $30.00over-272k
  • ultrafast·$12.00 / $60.00
    • Prompt above 272,000 input tokens·$24.00 / $90.00over-272k
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and cache and 1.5x output for the whole request. Cache writes cost 1.25x input for the default 30-minute entry and 2.5x for a 12-hour one (Responses API).
GPT-6.1 Terragpt-6.1-terra$1.00$5.001,050,0000.05x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1.25x for 30 min

write 2.5x for 12 h

Bands
  • Prompt above 272,000 input tokens·$2.00 / $7.50over-272k
Service tiers
  • flex·$0.5000 / $2.50
    • Prompt above 272,000 input tokens·$1.00 / $3.75over-272k
  • fast·$2.00 / $10.00
    • Prompt above 272,000 input tokens·$4.00 / $15.00over-272k
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and cache and 1.5x output for the whole request. Cache writes cost 1.25x input for the default 30-minute entry and 2.5x for a 12-hour one (Responses API).
GPT-6 Solgpt-6-sol$2.00$10.001,050,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1.25x for 30 min

write 2.5x for 12 h

Bands
  • Prompt above 272,000 input tokens·$4.00 / $15.00over-272k
Service tiers
  • flex·$1.00 / $5.00
    • Prompt above 272,000 input tokens·$2.00 / $7.50over-272k
  • fast·$4.00 / $20.00
    • Prompt above 272,000 input tokens·$8.00 / $30.00over-272k
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and cache and 1.5x output for the whole request. Cache writes cost 1.25x input for the default 30-minute entry and 2.5x for a 12-hour one (Responses API).
GPT-6 Lunagpt-6-luna$0.1000$0.50001,050,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1.25x for 30 min

write 2.5x for 12 h

Bands
  • Prompt above 272,000 input tokens·$0.2000 / $0.7500over-272k
Service tiers
  • flex·$0.0500 / $0.2500
    • Prompt above 272,000 input tokens·$0.1000 / $0.3750over-272k
  • fast·$0.2000 / $1.00
    • Prompt above 272,000 input tokens·$0.4000 / $1.50over-272k
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and cache and 1.5x output for the whole request. Cache writes cost 1.25x input for the default 30-minute entry and 2.5x for a 12-hour one (Responses API).
GPT-5.6 Solgpt-5.6-sol$4.00$20.001,050,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1.25x for 30 min

write 2.5x for 12 h

Bands
  • Prompt above 272,000 input tokens·$8.00 / $30.00over-272k
Service tiers
  • flex·$2.00 / $10.00
    • Prompt above 272,000 input tokens·$4.00 / $15.00over-272k
  • fast·$8.00 / $40.00
    • Prompt above 272,000 input tokens·$16.00 / $60.00over-272k
Also billed as
  • gpt-5.6
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and cache and 1.5x output for the whole request. Cache writes cost 1.25x input for the default 30-minute entry and 2.5x for a 12-hour one (Responses API). $4/$20 is promotional until 2026-11-30, and OpenAI has not published the price after it as a figure.
GPT-5.6 Terragpt-5.6-terra$2.00$12.001,050,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1.25x for 30 min

write 2.5x for 12 h

Bands
  • Prompt above 272,000 input tokens·$4.00 / $18.00over-272k
Service tiers
  • flex·$1.00 / $6.00
    • Prompt above 272,000 input tokens·$2.00 / $9.00over-272k
  • fast·$4.00 / $24.00
    • Prompt above 272,000 input tokens·$8.00 / $36.00over-272k
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and cache and 1.5x output for the whole request. Cache writes cost 1.25x input for the default 30-minute entry and 2.5x for a 12-hour one (Responses API).
GPT-5.6 Lunagpt-5.6-luna$0.2000$1.201,050,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1.25x for 30 min

write 2.5x for 12 h

Bands
  • Prompt above 272,000 input tokens·$0.4000 / $1.80over-272k
Service tiers
  • flex·$0.1000 / $0.6000
    • Prompt above 272,000 input tokens·$0.2000 / $0.9000over-272k
  • fast·$0.4000 / $2.40
    • Prompt above 272,000 input tokens·$0.8000 / $3.60over-272k
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and cache and 1.5x output for the whole request. Cache writes cost 1.25x input for the default 30-minute entry and 2.5x for a 12-hour one (Responses API).
GPT-5.6 Cybergpt-5.6-cyberNot offered as an alternative$12.50$75.00400,0000.1xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1.25x for 30 min

write 2.5x for 12 h

Catalogue note
Needs approval through the Daybreak programme. At most 272,000 input tokens, Responses API only, no batch. Cache writes cost 1.25x input for the default 30-minute entry and 2.5x for a 12-hour one.
GPT-5gpt-5Not offered as an alternative$1.25$10.00400,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • flex·$0.6250 / $5.00
  • fast·$2.50 / $20.00
Also billed as
  • gpt-5-2025-08-07
Catalogue note
Caching is automatic, with no cache_control to set. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. The 2025-08-07 snapshot shuts down on 2026-12-11; OpenAI names the gpt-5.6 models as replacements. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-5 minigpt-5-miniNot offered as an alternative$0.2500$2.00400,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • flex·$0.1250 / $1.00
  • fast·$0.4500 / $3.60
Also billed as
  • gpt-5-mini-2025-08-07
Catalogue note
Caching is automatic, with no cache_control to set. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. The 2025-08-07 snapshot shuts down on 2026-12-11; OpenAI names the gpt-5.6 models as replacements. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-5 nanogpt-5-nanoNot offered as an alternative$0.0500$0.4000400,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • flex·$0.0250 / $0.2000
Also billed as
  • gpt-5-nano-2025-08-07
Catalogue note
Caching is automatic, with no cache_control to set. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. The 2025-08-07 snapshot shuts down on 2026-12-11; OpenAI names the gpt-5.6 models as replacements. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-5.5gpt-5.5$5.00$30.001,050,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt above 272,000 input tokens·$10.00 / $45.00over-272k
Service tiers
  • flex·$2.50 / $15.00
    • Prompt above 272,000 input tokens·$5.00 / $22.50over-272k
  • fast·$12.50 / $75.00
Also billed as
  • gpt-5.5-2026-04-23
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and 1.5x output for the whole request. Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor.
GPT-5.5 progpt-5.5-pro$30.00$180.001,050,000-0.5x
Bands, tiers and notes
Cache

No prompt caching

Bands
  • Prompt above 272,000 input tokens·$60.00 / $270.00over-272k
Service tiers
  • flex·$15.00 / $90.00
Also billed as
  • gpt-5.5-pro-2026-04-23
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and 1.5x output for the whole request. Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor.
GPT-5.4gpt-5.4$2.50$15.001,050,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt above 272,000 input tokens·$5.00 / $22.50over-272k
Service tiers
  • flex·$1.25 / $7.50
    • Prompt above 272,000 input tokens·$2.50 / $11.25over-272k
  • fast·$5.00 / $30.00
Also billed as
  • gpt-5.4-2026-03-05
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and 1.5x output for the whole request. Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor.
GPT-5.4 progpt-5.4-pro$30.00$180.001,050,000-0.5x
Bands, tiers and notes
Cache

No prompt caching

Bands
  • Prompt above 272,000 input tokens·$60.00 / $270.00over-272k
Service tiers
  • flex·$15.00 / $90.00
    • Prompt above 272,000 input tokens·$30.00 / $135.00over-272k
Also billed as
  • gpt-5.4-pro-2026-03-05
Catalogue note
Priced per request by prompt length: over 272,000 input tokens, 2x input and 1.5x output for the whole request. Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor.
GPT-5.4 minigpt-5.4-mini$0.7500$4.50400,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • flex·$0.3750 / $2.25
  • fast·$1.50 / $9.00
Also billed as
  • gpt-5.4-mini-2026-03-17
Catalogue note
Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor.
GPT-5.4 nanogpt-5.4-nanoNot offered as an alternative$0.2000$1.25400,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • flex·$0.1000 / $0.6250
Also billed as
  • gpt-5.4-nano-2026-03-17
Catalogue note
Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-5.2gpt-5.2$1.75$14.00400,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • flex·$0.8750 / $7.00
  • fast·$3.50 / $28.00
Also billed as
  • gpt-5.2-2025-12-11
Catalogue note
Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor.
GPT-5.2 progpt-5.2-pro$21.00$168.00400,000-0.5x
Bands, tiers and notes
Cache

No prompt caching

Also billed as
  • gpt-5.2-pro-2025-12-11
Catalogue note
OpenAI publishes no cached-input price for this model, so caching is not priced.
GPT-5.1gpt-5.1Not offered as an alternative$1.25$10.00400,0000.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • flex·$0.6250 / $5.00
  • fast·$2.50 / $20.00
Also billed as
  • gpt-5.1-2025-11-13
Catalogue note
Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-5 progpt-5-proNot offered as an alternative$15.00$120.00400,000-0.5x
Bands, tiers and notes
Cache

No prompt caching

Also billed as
  • gpt-5-pro-2025-10-06
Catalogue note
OpenAI publishes no cached-input price for this model, so caching is not priced. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-4.1gpt-4.1Not offered as an alternative$2.00$8.001,047,5760.25x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • fast·$3.50 / $14.00
Also billed as
  • gpt-4.1-2025-04-14
Catalogue note
Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-4.1 minigpt-4.1-miniNot offered as an alternative$0.4000$1.601,047,5760.25x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • fast·$0.7000 / $2.80
Also billed as
  • gpt-4.1-mini-2025-04-14
Catalogue note
Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-4ogpt-4oNot offered as an alternative$2.50$10.00128,0000.5x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • fast·$4.25 / $17.00
Also billed as
  • gpt-4o-2024-11-20
  • gpt-4o-2024-08-06
Catalogue note
Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-4o minigpt-4o-miniNot offered as an alternative$0.1500$0.6000128,0000.5x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • fast·$0.2500 / $1.00
Also billed as
  • gpt-4o-mini-2024-07-18
Catalogue note
Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
o3o3Not offered as an alternative$2.00$8.00200,0000.25x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • flex·$1.00 / $4.00
  • fast·$3.50 / $14.00
Also billed as
  • o3-2025-04-16
Catalogue note
Caching is automatic; a write costs what input costs. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
o3-proo3-proNot offered as an alternative$20.00$80.00200,000-0.5x
Bands, tiers and notes
Cache

No prompt caching

Also billed as
  • o3-pro-2025-06-10
Catalogue note
OpenAI publishes no cached-input price for this model, so caching is not priced. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-5.3-Codexgpt-5.3-codexNot offered as an alternative$1.75$14.00400,0000.1xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Service tiers
  • fast·$3.50 / $28.00
Catalogue note
Caching is automatic; a write costs what input costs. Responses API only, and no batch. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. At most 272,000 input tokens. Deprecated by OpenAI: priced for the logs that hold it, never recommended.
GPT-5.2-Codexgpt-5.2-codexNot offered as an alternative$1.75$14.00400,0000.1xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Caching is automatic; a write costs what input costs. Responses API only, and no batch. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. At most 272,000 input tokens. Shut down by OpenAI: priced for the logs that hold it, never recommended.
GPT-5.1-Codexgpt-5.1-codexNot offered as an alternative$1.25$10.00400,0000.1xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Caching is automatic; a write costs what input costs. Responses API only, and no batch. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Shut down by OpenAI: priced for the logs that hold it, never recommended.
GPT-5.1-Codex-Maxgpt-5.1-codex-maxNot offered as an alternative$1.25$10.00400,0000.1xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Caching is automatic; a write costs what input costs. Responses API only, and no batch. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Shut down by OpenAI: priced for the logs that hold it, never recommended.
GPT-5.1-Codex Minigpt-5.1-codex-miniNot offered as an alternative$0.2500$2.00400,0000.1xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Caching is automatic; a write costs what input costs. Responses API only, and no batch. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. Shut down by OpenAI: priced for the logs that hold it, never recommended.
GPT-5-Codexgpt-5-codexNot offered as an alternative$1.25$10.00400,0000.1xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Caching is automatic; a write costs what input costs. Responses API only, and no batch. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. At most 272,000 input tokens. Shut down by OpenAI: priced for the logs that hold it, never recommended.
codex-mini-latestcodex-mini-latestNot offered as an alternative$1.50$6.00200,0000.25xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Caching is automatic; a write costs what input costs. Responses API only, and no batch. OpenAI says the minimum for models before GPT-5.6 varies by request settings, so 1,024 here is the nearest published floor. A cache read costs 0.25x input. Shut down by OpenAI: priced for the logs that hold it, never recommended.
Chat Latestchat-latestNot offered as an alternative$5.00$30.00400,0000.1xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Points at the model ChatGPT currently uses, so it is priced and never recommended. Caching is automatic; a write costs what input costs. At most 272,000 input tokens, and no batch.

Listed by a provider, not priced here

These ids are on a provider’s own page and are left out of the catalogue on purpose, each for the reason beside it. A call to one reads as unpriced. It is never given a figure borrowed from another model.

  • gpt-5.5-cyber-previewNo priceDaybreak programme only, and no model page to read a context window from
  • gpt-6-rosalindNo pricetrusted-access programme only, and no model page to read a context window from
  • gpt-rosalind-researchNo pricetrusted-access programme only, and no model page to read a context window from
  • gpt-rosalind-discoveryNo pricetrusted-access programme only, and no model page to read a context window from
  • gpt-6-astra-lawNo priceno model page to read a context window from, and a promotional price the page dates only "at least through December 31, 2026"
  • gpt-6-cyberNo priceDaybreak programme only, and over 272K input tokens the page publishes no cache-write price, which a row would have to invent
  • gpt-5-search-apiNo priceno model page to read a context window from

Google9 models

Prices last read 2026-11-16·ai.google.dev/gemini-api/docs/pricing

ModelInputper 1M tokensOutputper 1M tokensContextCache readBatch
Gemini 2.5 Progemini-2.5-proRetiredPromotional until 2026-12-31Repriced$1.50$7.501,048,5760.1x0.5x
Bands, tiers and notes
Retired

Refused by the provider on 2026-08-28

The provider answered: This model models/gemini-2.5-pro is no longer available to new users. Please update your code to use models/gemini-3.1-pro-preview for the latest features and improvements. We recommend you to use the Interactions API.

Promotional rate
$0.7500 / $3.75until 2026-12-31, inclusive

still stated by the provider’s page on 2026-11-16

Cache

Caching: automatic

from 2,048 tokens

write 1x for 5 min

write 1x for 1 h

Earlier rates
  • until 2026-12-01T00:00:00Z·$1.25 / $10.00

The rates in the row apply after the last earlier period.

Catalogue note
From 2026-12-01 Google routes this id to Gemini 3.8 Flash and bills it at 3.8 Flash rates. Before then it bills $1.25/$10, and $2.50/$15.00 for a prompt over 200k tokens; this row prices every prompt before then at the up-to-200k rate, because no log Trazum prices says how long each prompt was. Implicit caching is automatic; explicit caching is also billed for storage per hour.
Gemini 2.5 Flashgemini-2.5-flashRetiredRepriced$0.3000$2.501,048,5760.1x0.5x
Bands, tiers and notes
Retired

Refused by the provider on 2026-08-28

The provider answered: This model models/gemini-2.5-flash is no longer available to new users. Please update your code to use models/gemini-3.6-flash for the latest features and improvements. We recommend you to use the Interactions API.

Cache

Caching: automatic

from 2,048 tokens

write 1x for 5 min

write 1x for 1 h

Earlier rates
  • until 2026-12-01T00:00:00Z·$0.3000 / $2.50

The rates in the row apply after the last earlier period.

Catalogue note
From 2026-12-01 Google routes this id to Gemini 3.5 Flash-Lite and bills it at 3.5 Flash-Lite rates, which for text are the $0.30/$2.50 and $0.03 cached this row carries.
Gemini 3.6 Flashgemini-3.6-flashPromotional until 2026-12-31$1.50$7.501,048,5760.1x0.5x
Bands, tiers and notes
Promotional rate
$0.7500 / $3.75until 2026-12-31, inclusive

still stated by the provider’s page on 2026-11-16

Cache

Caching: automatic

from 4,096 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Implicit caching is automatic above 4,096 tokens. Explicit caching is also billed for storage: $0.50 per 1M tokens per hour through 2026-12-31, $1.00 after.
Gemini 3.8 Flashgemini-3.8-flashPromotional until 2026-12-31$1.50$7.501,048,5760.1x0.5x
Bands, tiers and notes
Promotional rate
$0.7500 / $3.75until 2026-12-31, inclusive

still stated by the provider’s page on 2026-11-16

Cache

Caching: automatic

from 4,096 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Implicit caching is automatic above 4,096 tokens. Explicit caching is also billed for storage: $0.50 per 1M tokens per hour through 2026-12-31, $1.00 after.
Gemini 3.1 Pro Previewgemini-3.1-pro-preview$2.00$12.001,048,5760.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 4,096 tokens

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt above 200,000 input tokens·$4.00 / $18.00over-200k
Also billed as
  • gemini-3.1-pro-preview-customtools
Catalogue note
Priced per request by prompt length: $2/$12 up to 200,000 input tokens and $4/$18 over it. Implicit caching is automatic above 4,096 tokens; explicit caching is also billed for storage, $4.50 per 1M tokens per hour.
Gemini 4 Carbongemini-4-carbonPromotional until 2026-11-30$4.00$20.001,048,5760.05x0.5x
Bands, tiers and notes
Promotional rate
$2.00 / $10.00until 2026-11-30, inclusive

still stated by the provider’s page on 2026-11-16

Cache

Caching: automatic

from 4,096 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Implicit caching is automatic above 4,096 tokens and is the only kind: Google does not support explicit caching on Gemini 4. In a multi-turn conversation the previous request is cached only when the new one adds at least 2,050 tokens. Google also sells a Flex tier at half this price and a Priority tier at 1.8 times it; a call on either is priced here at the standard rate.
Gemini 4 Argongemini-4-argonPromotional until 2026-11-30$4.00$20.001,048,5760.05x0.5x
Bands, tiers and notes
Promotional rate
$2.00 / $10.00until 2026-11-30, inclusive

still stated by the provider’s page on 2026-11-16

Cache

Caching: automatic

from 4,096 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Implicit caching is automatic above 4,096 tokens and is the only kind: Google does not support explicit caching on Gemini 4. In a multi-turn conversation the previous request is cached only when the new one adds at least 2,050 tokens. Google also sells a Flex tier at half this price and a Priority tier at 1.8 times it; a call on either is priced here at the standard rate.
Gemini 4 Hydrogengemini-4-hydrogen$1.50$7.501,048,5760.05x0.5x
Bands, tiers and notes
Cache

Caching: automatic

from 4,096 tokens

write 1x for 5 min

write 1x for 1 h

Catalogue note
Implicit caching is automatic above 4,096 tokens and is the only kind: Google does not support explicit caching on Gemini 4. In a multi-turn conversation the previous request is cached only when the new one adds at least 2,050 tokens. Google also sells a Flex tier at half this price and a Priority tier at 1.8 times it; a call on either is priced here at the standard rate.
Gemini 3.5 Flash-Litegemini-3.5-flash-lite$0.3000$2.501,048,5760.1x0.5x
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Catalogue note
Implicit caching is automatic, and no minimum prefix length is published for this model. Explicit caching is also billed for storage, $1.00 per 1M tokens per hour.

Moonshot AI7 models

Prices last read 2026-11-05·platform.kimi.ai/docs/pricing/chat

ModelInputper 1M tokensOutputper 1M tokensContextCache readBatch
Kimi K3.1 Prokimi-k3.1-pro$2.00$10.001,048,5760.1x0.6x
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 2x for 1 h

Catalogue note
Cache writes are billed as their own line: 1x input for a 5-minute entry, 2x for an hour. Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Kimi K3.1 Pro (high speed)kimi-k3.1-pro-highspeed$4.00$20.001,048,5760.05xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 2x for 1 h

Catalogue note
Cache writes are billed as their own line: 1x input for a 5-minute entry, 2x for an hour. No batch API for this model. Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Kimi K3kimi-k3$3.00$15.001,048,5760.1x0.6x
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 2x for 1 h

Catalogue note
Cache writes are billed as their own line: 1x input for a 5-minute entry, 2x for an hour. Batch at 60% since the batch page listed it (read 2026-11-05). Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Kimi K2.7 Codekimi-k2.7-codeNot offered as an alternative$0.9500$4.00262,1440.2x0.6x
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Kimi K2.7 Code (high speed)kimi-k2.7-code-highspeedNot offered as an alternative$1.90$8.00262,1440.2xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Kimi K2.6kimi-k2.6Not offered as an alternative$0.9500$4.00262,1440.16842105263157897x0.6x
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Kimi K2kimi-k2Not offered as an alternative$0.6000$2.50256,0000.1xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Catalogue note
Discontinued by Moonshot on 2026-05-25; priced at the last rate it published. No batch API. Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.

DeepSeek4 models

Prices last read 2026-11-05·api-docs.deepseek.com/quick_start/pricing

ModelInputper 1M tokensOutputper 1M tokensContextCache readBatch
DeepSeek V3deepseek-v3Retired$0.2700$1.10128,0000.1xNo batch API
Bands, tiers and notes
Retired

Refused by the provider on 2026-08-28

The provider answered: The supported API model names are deepseek-v4-pro, deepseek-v4-flash, and deepseek-v4-flash-vision-exp, but you passed deepseek-v3.

Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

DeepSeek V4 Flashdeepseek-v4-flashRepricedNot offered as an alternative$0.1500$0.60001,000,0000.02xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Bands
  • UTC hours 01, 02, 03, 06, 07, 08, 09, Monday to Friday·$0.3000 / $1.20peak
Earlier rates
  • until 2026-09-10T04:00:00Z·$0.2200 / $0.6600
    • UTC hours 01, 02, 03, 06, 07, 08, 09, Monday to Friday·$0.4400 / $1.32peak

The rates in the row apply after the last earlier period.

Catalogue note
Retired 2026-09-10: the id is still accepted, served by V4.1 Flash and billed at the deepseek-flash price. Maximum output 384K tokens. Peak excludes Chinese public holidays, which this table cannot express.
DeepSeek V4.1 Flashdeepseek-flash$0.1500$0.60001,000,0000.02xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Bands
  • UTC hours 01, 02, 03, 06, 07, 08, 09, Monday to Friday·$0.3000 / $1.20peak
Catalogue note
Maximum output 384K tokens. Peak excludes Chinese public holidays, which this table cannot express: on a weekday holiday it charges peak where DeepSeek charges off-peak.
DeepSeek V4 Prodeepseek-v4-pro$0.6600$1.981,000,0000.0333xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

from 1,024 tokens

write 1x for 5 min

write 1x for 1 h

Bands
  • UTC hours 01, 02, 03, 06, 07, 08, 09, Monday to Friday·$1.32 / $3.96peak
Catalogue note
Maximum output 384K tokens.

xAI9 models

Prices last read 2026-11-05·docs.x.ai/developers/models

ModelInputper 1M tokensOutputper 1M tokensContextCache readBatch
Grok 4.7grok-4.7$2.00$6.00500,0000.25xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt of 200,000 input tokens or more·$4.00 / $12.00from-200k
Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Grok 4.6grok-4.6$2.00$6.00500,0000.25xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt of 200,000 input tokens or more·$4.00 / $12.00from-200k
Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Grok 4.5grok-4.5$2.00$6.00500,0000.15xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt of 200,000 input tokens or more·$4.00 / $12.00from-200k
Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Grok 4.3grok-4.3$1.25$2.501,000,0000.16x0.8x
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt of 200,000 input tokens or more·$2.50 / $5.00from-200k
Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Grok 4.20 (reasoning)grok-4.20-0309-reasoning$1.25$2.501,000,0000.16x0.8x
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt of 200,000 input tokens or more·$2.50 / $5.00from-200k
Also billed as
  • grok-4.20
  • grok-4.20-reasoning
  • grok-4.20-0309
Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Grok 4.20 (non-reasoning)grok-4.20-0309-non-reasoning$1.25$2.501,000,0000.16x0.8x
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt of 200,000 input tokens or more·$2.50 / $5.00from-200k
Also billed as
  • grok-4.20-non-reasoning
Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Grok 4.20 Multi-Agentgrok-4.20-multi-agent-0309$1.25$2.501,000,0000.16x0.8x
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt of 200,000 input tokens or more·$2.50 / $5.00from-200k
Also billed as
  • grok-4.20-multi-agent
Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Grok Build 0.1grok-build-0.1$1.00$2.00256,0000.2xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt of 200,000 input tokens or more·$2.00 / $4.00from-200k
Catalogue note
Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.
Grok 4grok-4RepricedNot offered as an alternative$1.25$2.50256,0000.16xNo batch API
Bands, tiers and notes
Cache

Caching: automatic

write 1x for 5 min

write 1x for 1 h

Bands
  • Prompt of 200,000 input tokens or more·$2.50 / $5.00from-200k
Earlier rates
  • until 2026-05-15T19:00:00Z·$3.00 / $15.00

The rates in the row apply after the last earlier period.

Also billed as
  • grok-4-0709
Catalogue note
Retired by xAI on 2026-05-15 and redirected to Grok 4.3, whose price it bills at from then; $3/$15 before. Caches automatically, but no minimum prefix length is published, so no caching advice is given for it.

Mistral AI4 models

Prices last read 2026-11-05·mistral.ai/pricing

ModelInputper 1M tokensOutputper 1M tokensContextCache readBatch
Mistral Large 2mistral-large-2Retired$2.00$6.00128,000-0.5x
Bands, tiers and notes
Retired

Refused by the provider on 2026-08-28

The provider answered: Invalid model: mistral-large-2

Cache

No prompt caching

Catalogue note
No prompt caching: reordering still helps readability but saves nothing here.
Mistral Large 3mistral-large-2512$0.5000$1.50262,144-0.5x
Bands, tiers and notes
Cache

Caching not known

Also billed as
  • mistral-large-latest
Catalogue note
Cached input costs 10% of the input price (mistral.ai/cloud/inference, read 2026-10-11). How caching starts and its minimum are not published anywhere Trazum can read, so no caching advice is given for it.
Mistral Medium 3.5mistral-medium-2604$1.50$7.50262,144-0.5x
Bands, tiers and notes
Cache

Caching not known

Also billed as
  • mistral-medium-latest
  • mistral-medium-3-5
Catalogue note
Cached input costs 10% of the input price (mistral.ai/cloud/inference, read 2026-10-11). How caching starts and its minimum are not published anywhere Trazum can read, so no caching advice is given for it.
Mistral Small 4mistral-small-2603$0.1500$0.6000262,144-0.5x
Bands, tiers and notes
Cache

Caching not known

Also billed as
  • mistral-small-latest
Catalogue note
Cached input costs 10% of the input price (mistral.ai/cloud/inference, read 2026-10-11). How caching starts and its minimum are not published anywhere Trazum can read, so no caching advice is given for it.

How to read a row

Retired: the provider refused a real request for the id, so it keeps its price for the logs that used it and is never recommended. Promotional: a time-limited rate, with its end date. No price: the catalogue holds none, and the page shows none.

RetiredPromotional until …No priceNot offered as an alternative

Where each price was read

A price is read from the provider’s page on a date, never remembered or worked out. These are the pages the catalogue names, and the date it last read each provider.