What this edition covers

Every row is a model the catalog listed as current (not legacy) on October 3, 2026, and every one was released by September 2026, so each was on sale on the last day of the quarter. The catalog's first verification came three days after the quarter closed. Its notes date the price moves it knows of (GPT-5.6 Sol's cut to $4/$20 on August 21; Google's promotional Flash rates, which run to December 31), none falls between September 30 and October 3, 2026, and every rate is unchanged in each later verification through October 31, 2026. On that record these are the prices that applied on September 30. Promotional prices still running at quarter end are the prices tabled, with the end date the vendor gives in the notes under each table.

xAI, DeepSeek and Mistral are not in this edition: the catalog does not cover them, and their rates were last read from their pages on June 21, before the quarter began. Reporting those as the quarter's prices would be a claim this page cannot check, so they are left out.

One catalog figure was corrected between the two verifications: Claude Sonnet 5.5's cache read was read as $0.20 on October 3, 2026 and verified as $0.10 (0.05x input) on October 9, 2026; the table uses the verified figure.

Prices are US dollars per million tokens (MTok) of text input and output at standard, real-time rates. "Cache write" is the vendor's cache-write rate where the catalog lists one (for Anthropic, the five-minute write); Google bills caching through the cache-read rate plus an hourly storage fee. A model marked † is still on sale but is not in its vendor's current lineup. The catalog does not record batch discounts, so none are tabled. The rows, with their source pages and verification dates, download as data.csv and data.json.

What changed during the quarter

  • Gemini 3.6 Flash (Google), released July 2026: $0.75 in, $3.75 out.
  • Gemini 3.5 Flash-Lite (Google), released July 2026: $0.30 in, $2.50 out.
  • GPT-5.6 Sol (OpenAI), released July 9, 2026: $4.00 in, $20.00 out.
  • GPT-5.6 Terra (OpenAI), released July 9, 2026: $2.00 in, $12.00 out.
  • GPT-5.6 Luna (OpenAI), released July 9, 2026: $0.20 in, $1.20 out.
  • Gemini 3.7 Flash (Google), released August 2026: $0.75 in, $3.75 out.
  • Gemini 3.8 Flash (Google), released September 2026: $0.75 in, $3.75 out.
  • Claude Fable 5.1 (Anthropic), released September 1, 2026: $10.00 in, $50.00 out.
  • GPT-6 Astra (OpenAI), released September 3, 2026: $10.00 in, $50.00 out.
  • Claude Opus 5.5 (Anthropic), released September 22, 2026: $4.00 in, $20.00 out.
  • GPT-6 Sol (OpenAI), released September 22, 2026: $2.00 in, $10.00 out.
  • GPT-6 Luna (OpenAI), released September 22, 2026: $0.10 in, $0.50 out.
  • Claude Sonnet 5.5 (Anthropic), released September 28, 2026: $2.00 in, $10.00 out.
  • GPT-6.1 Sol (OpenAI), released September 29, 2026: $2.00 in, $10.00 out.

The Q2 2026 edition priced the GPT-5.4, Claude Opus 4.6 and 4.7 and Gemini 2.5 generation on April 23, 2026.

Prices on September 30, 2026

Anthropic1

Model Released Input / MTok Output / MTok Cache read / MTok Cache write / MTok Context Max output
Claude Fable 5.1 Sep 1 $10.00 $50.00 $0.25 $12.50 1M 128K
Claude Opus 5.5 Sep 22 $4.00 $20.00 $0.20 $5.00 1M 128K
Claude Sonnet 5.5 Sep 28 $2.00 $10.00 $0.10 $2.50 1M 128K
Claude Haiku 4.5 Oct 2025 $1.00 $5.00 $0.10 $1.25 200K 64K
  • Claude Fable 5.1: Cache reads at 0.025x input (Claude Opus 5.5 and Claude Sonnet 5.5 0.05x, Claude Haiku 4.5 0.1x). Released September 1, 2026.
  • Claude Opus 5.5: Cache reads at 0.05x input. Fast mode (research preview) $8/$40. Released September 22, 2026.
  • Claude Sonnet 5.5: Released September 28, 2026.

Google2

Model Released Input / MTok Output / MTok Cache read / MTok Cache write / MTok Context Max output
Gemini 3.8 Flash Sep 2026 $0.75 $3.75 $0.075 — 1,048,576 65,536
Gemini 3.7 Flash † Aug 2026 $0.75 $3.75 $0.075 — 1,048,576 65,536
Gemini 3.6 Flash † Jul 2026 $0.75 $3.75 $0.075 — 1,048,576 65,536
Gemini 3.5 Flash † May 2026 $1.50 $9.00 $0.15 — 1,048,576 65,536
Gemini 3.5 Flash-Lite Jul 2026 $0.30 $2.50 $0.03 — 1,048,576 65,536
Gemini 3.1 Pro Preview Feb 2026 $2.00 $12.00 $0.20 — 1,048,576 65,536
  • Gemini 3.8 Flash: Promotional $0.75/$3.75 through December 31, 2026, then $1.50/$7.50. Context caching is per cached token plus an hourly storage fee. Batch/Flex 50% off.
  • Gemini 3.7 Flash: Same promotional pricing as 3.8 Flash through December 31, 2026. Context caching is per cached token plus an hourly storage fee. Batch/Flex 50% off.
  • Gemini 3.6 Flash: Same promotional pricing as 3.8 Flash through December 31, 2026. Context caching is per cached token plus an hourly storage fee. Batch/Flex 50% off.
  • Gemini 3.5 Flash: Context caching is per cached token plus an hourly storage fee. Batch/Flex 50% off.
  • Gemini 3.5 Flash-Lite: Replaces Gemini 3.1 Flash-Lite (shuts down May 7, 2027). Context caching is per cached token plus an hourly storage fee. Batch/Flex 50% off.
  • Gemini 3.1 Pro Preview: Prompts over 200K tokens: $4/$18. Preview model. Context caching is per cached token plus an hourly storage fee. Batch/Flex 50% off.

OpenAI34567891011121314

Model Released Input / MTok Output / MTok Cache read / MTok Cache write / MTok Context Max output
GPT-6 Astra Sep 3 $10.00 $50.00 $1.00 $12.50 1.05M 128K
GPT-6.1 Sol Sep 29 $2.00 $10.00 $0.10 $2.50 1.05M 128K
GPT-6 Sol † Sep 22 $2.00 $10.00 $0.20 $2.50 1.05M 128K
GPT-6 Luna Sep 22 $0.10 $0.50 $0.01 $0.125 1.05M 128K
GPT-5.6 Sol † Jul 9 $4.00 $20.00 $0.40 $5.00 1.05M 128K
GPT-5.6 Terra † Jul 9 $2.00 $12.00 $0.20 $2.50 1.05M 128K
GPT-5.6 Luna † Jul 9 $0.20 $1.20 $0.02 $0.25 1.05M 128K
GPT-5.5 Apr 2026 $5.00 $30.00 $0.50 — 1.05M 128K
GPT-5.5 Pro † Apr 2026 $30.00 $180.00 — — 1.05M 128K
GPT-5.4 † Mar 2026 $2.50 $15.00 $0.25 — 1.05M 128K
GPT-5.4 Pro † Mar 2026 $30.00 $180.00 — — 1.05M 128K
GPT-5.4 Mini Mar 2026 $0.75 $4.50 $0.075 — 400K 128K
  • GPT-6 Astra: OpenAI's most capable model. Released September 3, 2026. Prompts over 272K input tokens bill 2x input and cache rates and 1.5x output for the whole request. Cache writes 1.25x input.
  • GPT-6.1 Sol: Released September 29, 2026. Prompts over 272K input tokens bill 2x input and cache rates and 1.5x output for the whole request. Cache writes 1.25x input. Cached input is 5% of input.
  • GPT-6 Sol: Superseded by GPT-6.1 Sol at the same input and output price. Released September 22, 2026. Prompts over 272K input tokens bill 2x input and cache rates and 1.5x output for the whole request. Cache writes 1.25x input.
  • GPT-6 Luna: Released September 22, 2026. Prompts over 272K input tokens bill 2x input and cache rates and 1.5x output for the whole request. Cache writes 1.25x input.
  • GPT-5.6 Sol: Released July 9, 2026; cut to $4/$20 on Aug 21, 2026 (promotional pricing through Nov 21, 2026). Cache writes 1.25x input. Prompts over 272K input tokens bill 2x input / 1.5x output for the whole request.
  • GPT-5.6 Terra: Released July 9, 2026. Prompts over 272K input tokens bill 2x input / 1.5x output for the whole request.
  • GPT-5.6 Luna: Released July 9, 2026. Prompts over 272K input tokens bill 2x input / 1.5x output for the whole request.
  • GPT-5.5: Released April 24, 2026. Prompts over 272K input tokens bill 2x input / 1.5x output for the whole request.
  • GPT-5.5 Pro: No cached-input price published.
  • GPT-5.4: Released March 5, 2026. Prompts over 272K input tokens bill 2x input / 1.5x output for the whole request.
  • GPT-5.4 Mini: Released March 17, 2026.

† On sale at quarter end, not in the vendor's current lineup.

Scenario A: Nightly SEC-filing triage

500 10-Ks a night, 4 fact-extraction questions per filing, 50,000 input tokens per question (the filing plus the extraction prompt) and 1,000 output tokens per answer, over 21 trading days: 42,000 queries, 2,100 MTok in and 42 MTok out a month.

Model Input Output Cost per month
GPT-6 Luna $210.00 $21.00 $231.00
Gemini 3.5 Flash-Lite $630.00 $105.00 $735.00
Gemini 3.8 Flash $1,575.00 $157.50 $1,732.50
GPT-5.4 Mini $1,575.00 $189.00 $1,764.00
Claude Haiku 4.5 $2,100.00 $210.00 $2,310.00
Claude Sonnet 5.5 $4,200.00 $420.00 $4,620.00
GPT-6.1 Sol $4,200.00 $420.00 $4,620.00
Gemini 3.1 Pro Preview $4,200.00 $504.00 $4,704.00
Claude Opus 5.5 $8,400.00 $840.00 $9,240.00
GPT-5.5 $10,500.00 $1,260.00 $11,760.00
Claude Fable 5.1 $21,000.00 $2,100.00 $23,100.00
GPT-6 Astra $21,000.00 $2,100.00 $23,100.00

Scenario B: Quarterly earnings-call summaries

20 tickers, one call each per quarter, 50,000 input tokens per call (transcript plus context) and 3,000 output tokens: 1 MTok in and 0.06 MTok out a quarter.

Model Input Output Cost per quarter
GPT-6 Luna $0.10 $0.03 $0.13
Gemini 3.5 Flash-Lite $0.30 $0.15 $0.45
Gemini 3.8 Flash $0.75 $0.22 $0.98
GPT-5.4 Mini $0.75 $0.27 $1.02
Claude Haiku 4.5 $1.00 $0.30 $1.30
Claude Sonnet 5.5 $2.00 $0.60 $2.60
GPT-6.1 Sol $2.00 $0.60 $2.60
Gemini 3.1 Pro Preview $2.00 $0.72 $2.72
Claude Opus 5.5 $4.00 $1.20 $5.20
GPT-5.5 $5.00 $1.80 $6.80
Claude Fable 5.1 $10.00 $3.00 $13.00
GPT-6 Astra $10.00 $3.00 $13.00

Scenario C: Multi-filing peer comparison

5 filings in one call, 125,000 input tokens and 2,000 output tokens per run, daily over 21 business days: 2.625 MTok in and 0.042 MTok out a month.

Model Input Output Cost per month
GPT-6 Luna $0.26 $0.02 $0.28
Gemini 3.5 Flash-Lite $0.79 $0.11 $0.89
Gemini 3.8 Flash $1.97 $0.16 $2.13
GPT-5.4 Mini $1.97 $0.19 $2.16
Claude Haiku 4.5 $2.63 $0.21 $2.84
Claude Sonnet 5.5 $5.25 $0.42 $5.67
GPT-6.1 Sol $5.25 $0.42 $5.67
Gemini 3.1 Pro Preview $5.25 $0.50 $5.75
Claude Opus 5.5 $10.50 $0.84 $11.34
GPT-5.5 $13.13 $1.26 $14.39
Claude Fable 5.1 $26.25 $2.10 $28.35
GPT-6 Astra $26.25 $2.10 $28.35

Scenario D: Intraday research loop

10 ideas a day, 8,000 input tokens and 1,000 output tokens per idea, over 30 days: 2.4 MTok in and 0.3 MTok out a month.

Model Input Output Cost per month
GPT-6 Luna $0.24 $0.15 $0.39
Gemini 3.5 Flash-Lite $0.72 $0.75 $1.47
Gemini 3.8 Flash $1.80 $1.13 $2.93
GPT-5.4 Mini $1.80 $1.35 $3.15
Claude Haiku 4.5 $2.40 $1.50 $3.90
Claude Sonnet 5.5 $4.80 $3.00 $7.80
GPT-6.1 Sol $4.80 $3.00 $7.80
Gemini 3.1 Pro Preview $4.80 $3.60 $8.40
Claude Opus 5.5 $9.60 $6.00 $15.60
GPT-5.5 $12.00 $9.00 $21.00
Claude Fable 5.1 $24.00 $15.00 $39.00
GPT-6 Astra $24.00 $15.00 $39.00

Every scenario prompt stays under the long-prompt boundaries the catalog records (200K tokens on Gemini 3.1 Pro Preview, 272K on OpenAI's GPT-5.4 and later), so the base rates apply. Caching changes scenario A most: the four questions on each filing share the same 50,000-token filing, and a cached prefix bills at the cache-read rate on the second to fourth question. The Token-Cost Optimizer and the Financial Document Token Estimator take the same inputs and price a workload of your own on the current catalog.

Cheapest and dearest per workload

Lowest and highest cost at standard rates across the 12 lineup models. Cheapest is not best: output quality is not measured here, and a triage model that misreads a filing costs more than its tokens.

Workload Cheapest Runner-up Dearest
Nightly SEC-filing triage GPT-6 Luna ($231.00) Gemini 3.5 Flash-Lite ($735.00) GPT-6 Astra ($23,100.00)
Quarterly earnings-call summaries GPT-6 Luna ($0.13) Gemini 3.5 Flash-Lite ($0.45) GPT-6 Astra ($13.00)
Multi-filing peer comparison GPT-6 Luna ($0.28) Gemini 3.5 Flash-Lite ($0.89) GPT-6 Astra ($28.35)
Intraday research loop GPT-6 Luna ($0.39) Gemini 3.5 Flash-Lite ($1.47) GPT-6 Astra ($39.00)

Limits

This is a price inventory for one date, not a ranking. Choosing a model also takes output quality on the task, latency and rate limits, none of which are measured here; the Hallucination Detector and the Price-Blind Auditor cover the checking side. Enterprise contracts differ from the self-serve rates tabled. Prices have moved since the quarter ended (new models launched in October, and announced changes are dated in the catalog), so a spending decision should use the current prices rather than this record.

References

  1. Anthropic. platform.claude.com/docs/en/about-claude/pricing (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  2. Google. ai.google.dev/gemini-api/docs/pricing (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  3. OpenAI. developers.openai.com/api/docs/models/gpt-6-astra (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  4. OpenAI. developers.openai.com/api/docs/models/gpt-6.1-sol (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  5. OpenAI. developers.openai.com/api/docs/models/gpt-6-sol (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  6. OpenAI. developers.openai.com/api/docs/models/gpt-6-luna (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  7. OpenAI. developers.openai.com/api/docs/models/gpt-5.6-sol (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  8. OpenAI. developers.openai.com/api/docs/models/gpt-5.6-terra (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  9. OpenAI. developers.openai.com/api/docs/models/gpt-5.6-luna (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  10. OpenAI. developers.openai.com/api/docs/models/gpt-5.5 (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  11. OpenAI. developers.openai.com/api/docs/models/gpt-5.5-pro (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  12. OpenAI. developers.openai.com/api/docs/models/gpt-5.4 (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  13. OpenAI. developers.openai.com/api/docs/models/gpt-5.4-pro (verified 2026-10-09, via the aidevhub.io model catalog). ↩

  14. OpenAI. developers.openai.com/api/docs/models/gpt-5.4-mini (verified 2026-10-09, via the aidevhub.io model catalog). ↩