AI models compared
Pick two to four models from Anthropic, OpenAI, Google. See their price, context window, and retirement status side by side, and what your workload costs on each. Each model page links every fact to the vendor's own page.
At 10,000 requests a month of 5,000 input and 500 output tokens, the cheapest model here is GPT-6 Luna at about $7.50 a month. The most expensive is Claude Fable 5.1 at about $975. Google lists May 7, 2027 as the earliest possible retirement date for Gemini 3.1 Flash-Lite, and says it will send the exact date later. Google names Gemini 3.5 Flash-Lite as the replacement. Checked 2026-10-21.
Pick 2 to 4 models
3 of 4 picked.
Prices are USD per 1M tokens, standard tier, checked 2026-10-21. Monthly cost uses the workload below.
Claude Sonnet 5.5 Anthropic
- Input per 1M
- $2.00
- Output per 1M
- $10
- Monthly cost
- $195
- $20 per 1,000 requests
- Context
- 1M
- Retirement
- Active
- Anthropic says not retired before Sep 28, 2027
- API ID:
- claude-sonnet-5-5
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Sep 28, 2026
- Knowledge cutoff:
- Jun 2026
GPT-6.1 Sol OpenAI
- Input per 1M
- $2.00
- Output per 1M
- $10
- Monthly cost
- $150
- $15 per 1,000 requests
- Context
- 1.05M
- Retirement
- Active
- No retirement date listed
- API ID:
- gpt-6.1-sol
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Not published
- Knowledge cutoff:
- Apr 30, 2026
Gemini 3.8 Flash Google
- Input per 1M
- $0.75
- Output per 1M
- $3.75
- Monthly cost
- $56Cheapest
- $5.63 per 1,000 requests
- Context
- 1,048,576
- Retirement
- Active
- No retirement date listed
- API ID:
- gemini-3.8-flash
- Max output:
- 65,536
- Inputs:
- Text, Images, Video, Audio, PDF
- Released:
- Sep 2, 2026
- Knowledge cutoff:
- Not published
- From Jan 1, 2027:
- $1.50 in, $7.50 out
| Model | Input per 1M | Output per 1M | Monthly cost | Retirement | Context |
|---|---|---|---|---|---|
| Claude Sonnet 5.5 Anthropic | $2.00 | $10 | $195 $20 per 1,000 requests | Active Anthropic says not retired before Sep 28, 2027 | 1M |
| API ID: claude-sonnet-5-5Max output: 128KInputs: Text, ImagesReleased: Sep 28, 2026Knowledge cutoff: Jun 2026 | |||||
| GPT-6.1 Sol OpenAI | $2.00 | $10 | $150 $15 per 1,000 requests | Active No retirement date listed | 1.05M |
| API ID: gpt-6.1-solMax output: 128KInputs: Text, ImagesReleased: Not publishedKnowledge cutoff: Apr 30, 2026 | |||||
| Gemini 3.8 Flash Google | $0.75 | $3.75 | $56Cheapest $5.63 per 1,000 requests | Active No retirement date listed | 1,048,576 |
| API ID: gemini-3.8-flashMax output: 65,536Inputs: Text, Images, Video, Audio, PDFReleased: Sep 2, 2026Knowledge cutoff: Not publishedFrom Jan 1, 2027: $1.50 in, $7.50 out | |||||
Monthly cost estimate
Example sizes are our round numbers. Set your own below.
Claude Sonnet 5.5 $195 a month
$20 per 1,000 requests
GPT-6.1 Sol $150 a month
$15 per 1,000 requests
Gemini 3.8 Flash Cheapest $56 a month
$5.63 per 1,000 requests
List prices checked 2026-10-21, standard tier, no caching, no batch discount. Claude 4.7 and later count about 30% more tokens for the same text, so their rows include that. Long-prompt prices apply where the vendor sets them.
Which model scores higher? Independent benchmarks
Epoch Capabilities Index: Claude Opus 5.5 scores highest of the 17 models here. GPT-6 Astra scores 168.81 against 168.83, well inside Epoch's range, so treat the two as level. GPQA Diamond: Gemini 4 Argon scores highest of the 16 models here. Both charts use scores from Epoch AI, which scores models from every vendor the same way. Checked 2026-10-21.
Epoch Capabilities Index
One score that combines many benchmarks, such as math, science, and coding tests, into a single scale. Epoch sets Claude 3.5 Sonnet at 130 and GPT-5 at 150.
-
1. Claude Opus 5.5 Anthropic 168.83
-
2. GPT-6 Astra OpenAI 168.81
-
3. GPT-6.1 Sol OpenAI 168.37
-
4. Claude Fable 5.1 Anthropic 166.49
-
5. Gemini 4 Argon Google 166.45
-
6. Claude Sonnet 5.5 Anthropic 166.18
-
7. GPT-5.6 Sol OpenAI 162.79
-
8. GPT-5.6 Terra OpenAI 160.09
-
9. Claude Haiku 5.5 Anthropic 159.32
-
10. Gemini 3.8 Flash Google 157.44
-
11. GPT-6 Luna OpenAI 156.75
-
12. GPT-5.6 Luna OpenAI 156.50
-
13. Gemini 3.1 Pro (preview) Google 155.34
-
14. Claude Sonnet 4.6 Anthropic 152.79
-
15. GPT-5.4 mini OpenAI 148.70
-
16. Gemini 3.5 Flash-Lite Google 144.91
-
17. Gemini 3.1 Flash-Lite Google 144.26
The dot is the score and the line is Epoch's range. The scale runs from 140 to 175, not from zero. Models whose lines overlap a lot are close.
GPQA Diamond
198 hard multiple-choice questions in biology, chemistry, and physics, written by PhD-level experts. Guessing scores 25%. We show the highest score for each model across the reasoning settings Epoch ran.
-
1. Gemini 4 Argon Google 96.4%
-
2. GPT-6 Astra OpenAI 95.8%
-
3. Claude Sonnet 5.5 Anthropic 95.6%
-
4. GPT-6.1 Sol OpenAI 95.4%
-
4. Gemini 3.8 Flash Google 95.4%
-
6. Gemini 3.1 Pro (preview) Google 94.4%
-
7. GPT-5.6 Sol OpenAI 93.5%
-
8. GPT-5.6 Terra OpenAI 93.3%
-
9. GPT-5.6 Luna OpenAI 91.6%
-
10. Claude Opus 5.5 Anthropic 90.6%
-
11. GPT-6 Luna OpenAI 90.5%
-
12. Claude Haiku 5.5 Anthropic 89.6%
-
13. Claude Sonnet 4.6 Anthropic 87.4%
-
14. GPT-5.4 mini OpenAI 86.9%
-
15. Gemini 3.5 Flash-Lite Google 83.3%
-
16. Gemini 3.1 Flash-Lite Google 81.8%
Each model's highest score across the reasoning settings Epoch ran. Epoch lists no score for Claude Fable 5.1.
Scores are Epoch AI's, not ours. Source: Epoch AI, Capabilities and benchmarking, from its public data download, checked 2026-10-21. Used under CC BY 4.0. We drew the charts. A benchmark is one test, so run your own inputs before you choose.
Every model in one table
Click a model for its full facts, sources, and a cost calculator. For shutdowns of older models, see our AI model retirement dates calendar. To get a shortlist for a job, try the AI model picker.
Prices are USD per 1M tokens, standard tier. Monthly cost is 10,000 requests of 5,000 input and 500 output tokens. Checked 2026-10-21 on each vendor's own pages.
Claude Fable 5.1 Anthropic
- Input per 1M
- $10
- Output per 1M
- $50
- Monthly cost
- $975
- $98 per 1,000 requests
- Context
- 1M
- Retirement
- Active
- Anthropic says not retired before Sep 1, 2027
- API ID:
- claude-fable-5-1
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Sep 1, 2026
- Knowledge cutoff:
- Jun 2026
Claude Opus 5.5 Anthropic
- Input per 1M
- $4.00
- Output per 1M
- $20
- Monthly cost
- $390
- $39 per 1,000 requests
- Context
- 1M
- Retirement
- Active
- Anthropic says not retired before Sep 22, 2027
- API ID:
- claude-opus-5-5
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Sep 22, 2026
- Knowledge cutoff:
- Jun 2026
Claude Sonnet 5.5 Anthropic
- Input per 1M
- $2.00
- Output per 1M
- $10
- Monthly cost
- $195
- $20 per 1,000 requests
- Context
- 1M
- Retirement
- Active
- Anthropic says not retired before Sep 28, 2027
- API ID:
- claude-sonnet-5-5
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Sep 28, 2026
- Knowledge cutoff:
- Jun 2026
Claude Haiku 5.5 Anthropic
- Input per 1M
- $0.10
- Output per 1M
- $0.50
- Monthly cost
- $9.75
- $0.98 per 1,000 requests
- Context
- 1M
- Retirement
- Active
- Anthropic says not retired before Oct 7, 2027
- API ID:
- claude-haiku-5-5
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Oct 7, 2026
- Knowledge cutoff:
- Jun 2026
Claude Sonnet 4.6 Anthropic
- Input per 1M
- $3.00
- Output per 1M
- $15
- Monthly cost
- $225
- $23 per 1,000 requests
- Context
- 1M
- Retirement
- Active
- Anthropic says not retired before Feb 17, 2027
- API ID:
- claude-sonnet-4-6
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Feb 17, 2026
- Knowledge cutoff:
- Aug 2025
GPT-6 Astra OpenAI
- Input per 1M
- $10
- Output per 1M
- $50
- Monthly cost
- $750
- $75 per 1,000 requests
- Context
- 1.05M
- Retirement
- Active
- No retirement date listed
- API ID:
- gpt-6-astra
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Not published
- Knowledge cutoff:
- Apr 30, 2026
GPT-6.1 Sol OpenAI
- Input per 1M
- $2.00
- Output per 1M
- $10
- Monthly cost
- $150
- $15 per 1,000 requests
- Context
- 1.05M
- Retirement
- Active
- No retirement date listed
- API ID:
- gpt-6.1-sol
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Not published
- Knowledge cutoff:
- Apr 30, 2026
GPT-6 Luna OpenAI
- Input per 1M
- $0.10
- Output per 1M
- $0.50
- Monthly cost
- $7.50Cheapest
- $0.75 per 1,000 requests
- Context
- 1.05M
- Retirement
- Active
- No retirement date listed
- API ID:
- gpt-6-luna
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Not published
- Knowledge cutoff:
- May 18, 2026
GPT-5.6 Sol OpenAI
- Input per 1M
- $4.00
- Output per 1M
- $20
- Monthly cost
- $300
- $30 per 1,000 requests
- Context
- 1.05M
- Retirement
- Active
- No retirement date listed
- API ID:
- gpt-5.6-sol
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Not published
- Knowledge cutoff:
- Feb 16, 2026
GPT-5.6 Terra OpenAI
- Input per 1M
- $2.00
- Output per 1M
- $12
- Monthly cost
- $160
- $16 per 1,000 requests
- Context
- 1.05M
- Retirement
- Active
- No retirement date listed
- API ID:
- gpt-5.6-terra
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Not published
- Knowledge cutoff:
- Feb 16, 2026
GPT-5.6 Luna OpenAI
- Input per 1M
- $0.20
- Output per 1M
- $1.20
- Monthly cost
- $16
- $1.60 per 1,000 requests
- Context
- 1.05M
- Retirement
- Active
- No retirement date listed
- API ID:
- gpt-5.6-luna
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Not published
- Knowledge cutoff:
- Feb 16, 2026
GPT-5.4 mini OpenAI
- Input per 1M
- $0.75
- Output per 1M
- $4.50
- Monthly cost
- $60
- $6.00 per 1,000 requests
- Context
- 400K
- Retirement
- Active
- No retirement date listed
- API ID:
- gpt-5.4-mini
- Max output:
- 128K
- Inputs:
- Text, Images
- Released:
- Not published
- Knowledge cutoff:
- Aug 31, 2025
Gemini 4 Argon Google
- Input per 1M
- $2.00
- Output per 1M
- $10
- Monthly cost
- $150
- $15 per 1,000 requests
- Context
- 1,048,576
- Retirement
- Active
- No retirement date listed
- API ID:
- gemini-4-argon
- Max output:
- 65,536
- Inputs:
- Text, Images, Video, Audio, PDF
- Released:
- Not published
- Knowledge cutoff:
- Not published
- From Dec 1, 2026:
- $4.00 in, $20 out
Gemini 3.1 Pro (preview) Google
- Input per 1M
- $2.00
- Output per 1M
- $12
- Monthly cost
- $160
- $16 per 1,000 requests
- Context
- 1,048,576
- Retirement
- Active
- No retirement date listed
- API ID:
- gemini-3.1-pro-preview
- Max output:
- 65,536
- Inputs:
- Text, Images, Video, Audio, PDF
- Released:
- Feb 19, 2026
- Knowledge cutoff:
- Not published
Gemini 3.8 Flash Google
- Input per 1M
- $0.75
- Output per 1M
- $3.75
- Monthly cost
- $56
- $5.63 per 1,000 requests
- Context
- 1,048,576
- Retirement
- Active
- No retirement date listed
- API ID:
- gemini-3.8-flash
- Max output:
- 65,536
- Inputs:
- Text, Images, Video, Audio, PDF
- Released:
- Sep 2, 2026
- Knowledge cutoff:
- Not published
- From Jan 1, 2027:
- $1.50 in, $7.50 out
Gemini 3.5 Flash-Lite Google
- Input per 1M
- $0.30
- Output per 1M
- $2.50
- Monthly cost
- $28
- $2.75 per 1,000 requests
- Context
- 1,048,576
- Retirement
- Active
- No retirement date listed
- API ID:
- gemini-3.5-flash-lite
- Max output:
- 65,536
- Inputs:
- Text, Images, Video, Audio, PDF
- Released:
- Jul 21, 2026
- Knowledge cutoff:
- Not published
Gemini 3.1 Flash-Lite Google
- Input per 1M
- $0.25
- Output per 1M
- $1.50
- Monthly cost
- $20
- $2.00 per 1,000 requests
- Context
- 1,048,576
- Retirement
- Earliest retirement May 7, 2027
- Google says this is the earliest possible date and it sends the exact date later
- API ID:
- gemini-3.1-flash-lite
- Max output:
- 65,536
- Inputs:
- Text, Images, Video, Audio, PDF
- Released:
- May 7, 2026
- Knowledge cutoff:
- Not published
| Model | Input per 1M | Output per 1M | Monthly cost | Retirement | Context |
|---|---|---|---|---|---|
| Claude Fable 5.1 Anthropic | $10 | $50 | $975 $98 per 1,000 requests | Active Anthropic says not retired before Sep 1, 2027 | 1M |
| API ID: claude-fable-5-1Max output: 128KInputs: Text, ImagesReleased: Sep 1, 2026Knowledge cutoff: Jun 2026 | |||||
| Claude Opus 5.5 Anthropic | $4.00 | $20 | $390 $39 per 1,000 requests | Active Anthropic says not retired before Sep 22, 2027 | 1M |
| API ID: claude-opus-5-5Max output: 128KInputs: Text, ImagesReleased: Sep 22, 2026Knowledge cutoff: Jun 2026 | |||||
| Claude Sonnet 5.5 Anthropic | $2.00 | $10 | $195 $20 per 1,000 requests | Active Anthropic says not retired before Sep 28, 2027 | 1M |
| API ID: claude-sonnet-5-5Max output: 128KInputs: Text, ImagesReleased: Sep 28, 2026Knowledge cutoff: Jun 2026 | |||||
| Claude Haiku 5.5 Anthropic | $0.10 | $0.50 | $9.75 $0.98 per 1,000 requests | Active Anthropic says not retired before Oct 7, 2027 | 1M |
| API ID: claude-haiku-5-5Max output: 128KInputs: Text, ImagesReleased: Oct 7, 2026Knowledge cutoff: Jun 2026 | |||||
| Claude Sonnet 4.6 Anthropic | $3.00 | $15 | $225 $23 per 1,000 requests | Active Anthropic says not retired before Feb 17, 2027 | 1M |
| API ID: claude-sonnet-4-6Max output: 128KInputs: Text, ImagesReleased: Feb 17, 2026Knowledge cutoff: Aug 2025 | |||||
| GPT-6 Astra OpenAI | $10 | $50 | $750 $75 per 1,000 requests | Active No retirement date listed | 1.05M |
| API ID: gpt-6-astraMax output: 128KInputs: Text, ImagesReleased: Not publishedKnowledge cutoff: Apr 30, 2026 | |||||
| GPT-6.1 Sol OpenAI | $2.00 | $10 | $150 $15 per 1,000 requests | Active No retirement date listed | 1.05M |
| API ID: gpt-6.1-solMax output: 128KInputs: Text, ImagesReleased: Not publishedKnowledge cutoff: Apr 30, 2026 | |||||
| GPT-6 Luna OpenAI | $0.10 | $0.50 | $7.50Cheapest $0.75 per 1,000 requests | Active No retirement date listed | 1.05M |
| API ID: gpt-6-lunaMax output: 128KInputs: Text, ImagesReleased: Not publishedKnowledge cutoff: May 18, 2026 | |||||
| GPT-5.6 Sol OpenAI | $4.00 | $20 | $300 $30 per 1,000 requests | Active No retirement date listed | 1.05M |
| API ID: gpt-5.6-solMax output: 128KInputs: Text, ImagesReleased: Not publishedKnowledge cutoff: Feb 16, 2026 | |||||
| GPT-5.6 Terra OpenAI | $2.00 | $12 | $160 $16 per 1,000 requests | Active No retirement date listed | 1.05M |
| API ID: gpt-5.6-terraMax output: 128KInputs: Text, ImagesReleased: Not publishedKnowledge cutoff: Feb 16, 2026 | |||||
| GPT-5.6 Luna OpenAI | $0.20 | $1.20 | $16 $1.60 per 1,000 requests | Active No retirement date listed | 1.05M |
| API ID: gpt-5.6-lunaMax output: 128KInputs: Text, ImagesReleased: Not publishedKnowledge cutoff: Feb 16, 2026 | |||||
| GPT-5.4 mini OpenAI | $0.75 | $4.50 | $60 $6.00 per 1,000 requests | Active No retirement date listed | 400K |
| API ID: gpt-5.4-miniMax output: 128KInputs: Text, ImagesReleased: Not publishedKnowledge cutoff: Aug 31, 2025 | |||||
| Gemini 4 Argon Google | $2.00 | $10 | $150 $15 per 1,000 requests | Active No retirement date listed | 1,048,576 |
| API ID: gemini-4-argonMax output: 65,536Inputs: Text, Images, Video, Audio, PDFReleased: Not publishedKnowledge cutoff: Not publishedFrom Dec 1, 2026: $4.00 in, $20 out | |||||
| Gemini 3.1 Pro (preview) Google | $2.00 | $12 | $160 $16 per 1,000 requests | Active No retirement date listed | 1,048,576 |
| API ID: gemini-3.1-pro-previewMax output: 65,536Inputs: Text, Images, Video, Audio, PDFReleased: Feb 19, 2026Knowledge cutoff: Not published | |||||
| Gemini 3.8 Flash Google | $0.75 | $3.75 | $56 $5.63 per 1,000 requests | Active No retirement date listed | 1,048,576 |
| API ID: gemini-3.8-flashMax output: 65,536Inputs: Text, Images, Video, Audio, PDFReleased: Sep 2, 2026Knowledge cutoff: Not publishedFrom Jan 1, 2027: $1.50 in, $7.50 out | |||||
| Gemini 3.5 Flash-Lite Google | $0.30 | $2.50 | $28 $2.75 per 1,000 requests | Active No retirement date listed | 1,048,576 |
| API ID: gemini-3.5-flash-liteMax output: 65,536Inputs: Text, Images, Video, Audio, PDFReleased: Jul 21, 2026Knowledge cutoff: Not published | |||||
| Gemini 3.1 Flash-Lite Google | $0.25 | $1.50 | $20 $2.00 per 1,000 requests | Earliest retirement May 7, 2027 Google says this is the earliest possible date and it sends the exact date later | 1,048,576 |
| API ID: gemini-3.1-flash-liteMax output: 65,536Inputs: Text, Images, Video, Audio, PDFReleased: May 7, 2026Knowledge cutoff: Not published | |||||
Common comparisons
- Claude Opus 5.5 vs GPT-6 Astra
- Claude Fable 5.1 vs GPT-6 Astra
- Claude Opus 5.5 vs Gemini 4 Argon
- Claude Sonnet 5.5 vs GPT-6.1 Sol
- Claude Sonnet 5.5 vs Gemini 3.8 Flash
- GPT-6.1 Sol vs Gemini 4 Argon
- Claude Haiku 5.5 vs GPT-6 Luna
- GPT-6 Luna vs Gemini 3.5 Flash-Lite
- Claude Sonnet 5.5 vs Claude Sonnet 4.6
- Gemini 3.5 Flash-Lite vs Gemini 3.1 Flash-Lite
- GPT-6.1 Sol vs GPT-5.6 Sol
How we check
- Every fact comes from the vendor's own model, pricing, or deprecation page. Each model page links the source next to each fact.
- If the vendor does not publish a fact, we write "Not published". We never fill a gap with a guess.
- Costs are list price, standard tier, no caching, no batch discount.
- Tiers follow where each vendor places the model in its own lineup. They are not a quality score.
- Benchmark scores come from Epoch AI, not from us or the vendors. We do not run benchmarks ourselves. Test the models on your own inputs before you choose.
Questions clients ask
Short answers, no sales language. If yours isn't here, ask an engineer directly.
Ask us anythingWhich AI model API is the cheapest?
Of the 17 models here, GPT-6 Luna costs the least at 10,000 requests a month of 5,000 input and 500 output tokens: about $7.50 a month. Claude Fable 5.1 costs the most, about $975. List prices, checked 2026-10-21.
Which current models are being retired?
Google lists May 7, 2027 as the earliest possible retirement date for Gemini 3.1 Flash-Lite, and says it will send the exact date later. Google names Gemini 3.5 Flash-Lite as the replacement. The other models on this page have no retirement date listed. Older models have their own dates in our retirement calendar.
Which AI model scores highest on benchmarks?
Epoch Capabilities Index: Claude Opus 5.5 scores highest of the 17 models here. GPT-6 Astra scores 168.81 against 168.83, well inside Epoch's range, so treat the two as level. GPQA Diamond: Gemini 4 Argon scores highest of the 16 models here. These are Epoch AI's scores, checked 2026-10-21. Top models sit within a few points of each other, so test them on your own inputs.
Why do Claude rows cost more in the estimator than the price per token suggests?
Anthropic says Claude 4.7 and later models count about 30% more tokens for the same text. The estimator adds that, so you compare the same text, not the same token count.
Where do these numbers come from?
From each vendor's own model pages, pricing pages, and deprecation pages, read on 2026-10-21. If a vendor does not publish a fact, the page says "Not published" instead of guessing.
Need a model in your product?
We test models on your real inputs and build the feature. We bill hourly or by milestone.