Terminal-Bench 4 (Vals AI) leaderboard
- Claude Opus 5.5 leads
- 65.2%
- models measured
- 48
- lead over #2
- +1.0 pts
- median
- 20.2%
Leader on Terminal-Bench 4 (Vals AI) · read October 21
# Terminal-Bench 4 (Vals AI), read October 21
leaders = {
"Claude Opus 5.5": 65.2%, "Claude Sonnet 5.5": 64.1%, "GPT-6 Astra": 59.6%,
}
On October 21, 2026, 48 of our catalog's models had a Terminal-Bench 4 (Vals AI) result, and Anthropic Claude Opus 5.5 led with 65.2%.
Tasks finished in a real terminal with one shared harness, run by Vals AI. Share completed.
Role in the index
A coding board we show beside the Vibe Coding Index, not inside it: the index keeps the four boards it was defined on, so a score does not move every time a board is added. Read the methodology
| # | Model | Result | Cost per test | Duration, as printed | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 65.2% | $13.20 | 1h 04m | 92.2 | $20.00 | |
| 2 | Claude Sonnet 5.5Anthropic | 64.1% | $16.51 | 1h 22m | 88.9 | $10.00 | |
| 3 | GPT-6 AstraOpenAI | 59.6% | $9.58 | 35m 42s | 88.7 | $50.00 | |
| 4 | Claude Fable 5.1Anthropic | 58.1% | $17.18 | 1h 00m | 89.1 | $50.00 | |
| 5 | Gemini 4 ArgonGoogle | 57.6% | $17.64 | 53m 35s | 80.5 | $10.00 | |
| 6 | GPT-6.1 SolOpenAI | 55.0% | $1.72 | 45m 12s | 88.7 | $10.00 | |
| 7 | Claude Opus 5Anthropic | 53.5% | $18.60 | 1h 07m | 81.8 | $25.00 | |
| 8 | Muse WonderMeta | 49.5% | $14.66 | 54m 41s | 64.5 | $20.00 | |
| 9 | GPT-6 SolOpenAI | 44.4% | $5.79 | 34m 56s | 82.8 | $10.00 | |
| 10 | Claude Fable 5Anthropic | 41.4% | $30.34 | 1h 08m | 82.1 | $50.00 | |
| 11 | GLM 5.3Z.ai | 38.9% | $9.37 | 1h 23m | 73.1 | $1.40 | |
| 12 | GPT-5.6 SolOpenAI | 37.9% | $7.98 | 33m 46s | 76.7 | $20.00 | |
| 13 | GLM 5.3 FlashZ.ai | 37.4% | $1.22 | 1h 01m | 71.0 | $0.50 | |
| 14 | Claude Haiku 5.5Anthropic | 36.4% | $3.12 | 51m 56s | 75.9 | $0.50 | |
| 15 | Qwen3.8 Max (0902)Qwen | 34.3% | $10.63 | 2h 28m | 63.5 | $6.00 | |
| 16 | DeepSeek V4.1 Flash 1017DeepSeek | 31.8% | $0.54 | 41m 10s | 83.3 | $0.60 | |
| 17 | Step 5 PreviewStepFun | 31.8% | $3.11 | 1h 58m | 71.4 | $2.70 | |
| 18 | MiMo-V2.6-ProXiaomi | 31.3% | $0.50 | 2h 42m | 72.6 | $0.87 | |
| 19 | Grok 4.7xAI | 28.8% | $18.01 | 50m 08s | 75.0 | $6.00 | |
| 20 | Mistral Large 4Mistral AI | 27.8% | $10.25 | 1h 40m | 55.3 | $4.18 | |
| 21 | MiMo-V2.6-FlashXiaomi | 24.2% | $0.22 | 2h 25m | 61.9 | $0.28 | |
| 22 | Claude Opus 4.8Anthropic | 23.2% | $17.14 | 1h 11m | 66.0 | $25.00 | |
| 23 | GPT-5.6 TerraOpenAI | 22.7% | $5.60 | 31m 29s | 69.0 | $12.00 | |
| 24 | Ember-1Fireworks | 20.7% | $19.48 | 1h 55m | Not benchmarked | $15.00 | |
| 25 | DeepSeek V4.1 FlashDeepSeek | 19.7% | $0.50 | 48m 22s | 77.2 | $0.525 | |
| 26 | Gemini 3.8 FlashGoogle | 19.7% | $8.65 | 1h 47m | 70.5 | $4.125 | |
| 27 | DeepSeek V4 Flash 0731DeepSeek | 18.7% | $0.77 | 1h 46m | Not benchmarked | $1.32 | |
| 28 | Kimi K3Moonshot AI | 17.2% | $12.48 | 3h 09m | 61.8 | $11.936 | |
| 29 | Grok 4.6xAI | 17.2% | $5.06 | 29m 25s | 76.4 | $6.00 | |
| 30 | DeepSeek V4 Pro 0813DeepSeek | 14.1% | $3.31 | 1h 32m | 54.6 | $1.98 | |
| 31 | GPT-6 LunaOpenAI | 13.6% | $0.35 | 32m 46s | 71.4 | $0.50 | |
| 32 | Gemini 3.7 FlashGoogle | 12.1% | $8.71 | 1h 24m | 69.9 | $3.75 | |
| 33 | GPT-5.6 LunaOpenAI | 11.6% | $0.73 | 35m 55s | 61.5 | $1.20 | |
| 34 | DeepSeek V4 Pro 0423DeepSeek | 11.1% | $3.45 | 1h 26m | 35.3 | $1.9105 | |
| 35 | Claude Sonnet 5Anthropic | 10.6% | $26.05 | 1h 45m | 63.5 | $10.00 | |
| 36 | Muse Spark 1.3Meta | 10.6% | $7.04 | 1h 59m | 79.7 | $4.25 | |
| 37 | Grok 4.5xAI | 8.6% | $5.52 | 33m 43s | 75.5 | $6.00 | |
| 38 | Hy4 previewTencent | 8.1% | $2.10 | 2h 10m | 61.6 | $2.501 | |
| 39 | Muse Spark OpenMeta | 6.6% | $7.26 | 2h 04m | 62.1 | $4.25 | |
| 40 | Gemini 3.5 FlashGoogle | 6.1% | $5.98 | 1h 11m | 42.8 | $9.00 | |
| 41 | Muse Spark 1.2Meta | 6.1% | $4.14 | 1h 18m | 65.1 | $4.25 | |
| 42 | Gemini 3.1 Pro PreviewGoogle | 2.5% | $4.11 | 18m 44s | 35.3 | $12.00 | |
| 43 | GPT-5.4 MiniOpenAI | 2.5% | $1.44 | 29m 30s | 28.2 | $4.50 | |
| 44 | MiniMax M3MiniMax | 1.0% | $7.23 | 1h 16m | 57.8 | $1.20 | |
| 45 | T InklingThinking Machines | 0.5% | $2.50 | 22m 59s | 49.7 | $4.05 | |
| 46 | MiMo-V2.5-ProXiaomi | 0.5% | $0.23 | 2h 07m | 40.0 | $0.609 | |
| 47 | Mercury 2.5Inception | 0.0% | $0.37 | 15m 14s | Not benchmarked | $0.15 | |
| 48 | T Inkling SmallThinking Machines | 0.0% | $0.32 | 11m 45s | 29.9 | $1.20 |
Only models in our catalog are listed, and each appears once with its best published run. 132 rows were read from the source on October 21, 2026, and rows that match no catalog model are left out. Cost per test is what Vals AI reports one test cost on the run behind the result, in US dollars. Every other column is a figure the board publishes for the same run.