65.15% · $16.00 per million tokens
Best AI model for coding
Ranked by Terminal-Bench 4.0: Frontier terminal and software-engineering tasks. Scores come from Vals AI (Mini-SWE-agent), not vendor self-reports. Claude Fable 5.2 leads at 65.15%.
35models scored
measured 2026-11-18
measured 2026-11-18
Coverage
Terminal-Bench 4.0 ranks 69 models; this table covers the 35 of them whose pricing we have verified against the provider's own page, so 34 scored models are not shown.
LeaderClaude Fable 5.2
Best score per dollarClaude Haiku 5.5
181.8 points per $ per million tokens
Top to bottom62.6pt
Ranked by Terminal-Bench 4.0
| Model | Maker | Score | Blended $/MTok | Score per $ |
|---|---|---|---|---|
| 1Claude Fable 5.2 | Anthropic | 65.15% | $16.00 | 4.1 |
| 2Claude Opus 5.5 | Anthropic | 65.15% | $8.00 | 8.1 |
| 3Claude Sonnet 5.5 | Anthropic | 64.14% | $4.00 | 16.0 |
| 4Kimi K3.1 Pro | Moonshot | 63.64% | $4.00 | 15.9 |
| 5gpt-6.2-astra | OpenAI | 63.13% | $16.00 | 3.9 |
| 6Gemini 4 Carbon | 63.13% | $4.00 | 15.8 | |
| 7Gemini 4 Neon | 61.62% | $6.00 | 10.3 | |
| 8gpt-6.2-sol | OpenAI | 60.61% | $4.00 | 15.2 |
| 9gpt-6-astra | OpenAI | 59.60% | $20.00 | 3.0 |
| 10Claude Fable 5.1 | Anthropic | 58.08% | $20.00 | 2.9 |
| 11Gemini 4 Argon | 57.58% | $4.00 | 14.4 | |
| 12gpt-6.1-sol | OpenAI | 55.05% | $4.00 | 13.8 |
| 13Gemini 4 Hydrogen | 54.04% | $3.00 | 18.0 | |
| 14Claude Opus 5 | Anthropic | 53.53% | $10.00 | 5.4 |
| 15gpt-6.1-terra | OpenAI | 49.49% | $2.00 | 24.7 |
| 16gpt-6-sol | OpenAI | 44.44% | $4.00 | 11.1 |
| 17Claude Fable 5 | Anthropic | 41.41% | $20.00 | 2.1 |
| 18GLM-5.3 | Z.ai | 38.89% | $2.15 | 18.1 |
| 19gpt-5.6-sol | OpenAI | 37.88% | $8.00 | 4.7 |
| 20GLM-5.3-Flash | Z.ai | 37.37% | $0.237 | 157.3 |
| 21Claude Haiku 5.5 | Anthropic | 36.36% | $0.2 | 181.8 |
| 22DeepSeek Flash V4.1 Flash 1017 | DeepSeek | 31.82% | $0.525 | 60.6 |
| 23Grok 4.7 | xAI | 28.79% | $3.00 | 9.6 |
| 24Claude Opus 4.8 | Anthropic | 23.23% | $10.00 | 2.3 |
| 25gpt-5.6-terra | OpenAI | 22.73% | $4.50 | 5.1 |
| 26Gemini 3.8 Flash | 19.70% | $1.50 | 13.1 | |
| 27Grok 4.6 | xAI | 17.17% | $3.00 | 5.7 |
| 28Kimi K3 | Moonshot | 17.17% | $6.00 | 2.9 |
| 29DeepSeek V4 Pro 0813 | DeepSeek | 14.14% | $1.98 | 7.1 |
| 30gpt-6-luna | OpenAI | 13.64% | $0.2 | 68.2 |
| 31gpt-5.6-luna | OpenAI | 11.62% | $0.45 | 25.8 |
| 32Claude Sonnet 5 | Anthropic | 10.61% | $4.00 | 2.7 |
| 33Grok 4.5 | xAI | 8.59% | $3.00 | 2.9 |
| 34gpt-5.4-mini | OpenAI | 2.52% | $1.69 | 1.5 |
| 35Gemini 3.1 Pro Preview Preview | 2.52% | $4.50 | 0.6 |
Cross-check: SWE-bench Verified
| Model | Maker | Score | Blended $/MTok | Score per $ |
|---|---|---|---|---|
| 1Claude Opus 5 | Anthropic | 97.00% | $10.00 | 9.7 |
| 2DeepSeek V4 Pro 0813 | DeepSeek | 96.40% | $1.98 | 48.7 |
| 3gpt-5.6-sol | OpenAI | 96.20% | $8.00 | 12.0 |
| 4Grok 4.6 | xAI | 95.60% | $3.00 | 31.9 |
| 5gpt-5.6-terra | OpenAI | 95.40% | $4.50 | 21.2 |
| 6GLM-5.3 | Z.ai | 95.40% | $2.15 | 44.4 |
| 7Claude Fable 5 | Anthropic | 95.00% | $20.00 | 4.8 |
| 8Kimi K3 | Moonshot | 93.40% | $6.00 | 15.6 |
| 9gpt-5.6-luna | OpenAI | 93.00% | $0.45 | 206.7 |
| 10GLM-5.3-Flash | Z.ai | 92.00% | $0.237 | 387.4 |
| 11Claude Opus 4.8 | Anthropic | 88.60% | $10.00 | 8.9 |
| 12Grok 4.5 | xAI | 86.60% | $3.00 | 28.9 |
| 13GLM-5.2 | Z.ai | 82.80% | $2.15 | 38.5 |
| 14gpt-5.5 | OpenAI | 82.60% | $11.25 | 7.3 |
| 15Claude Opus 4.7 | Anthropic | 82.00% | $10.00 | 8.2 |