AI Battle > Rankings > Terminal-Bench 4.0: Score
GPT-6 Astra (xhigh) by OpenAI leads the Terminal-Bench 4.0: Score ranking with a score of 0.60. The ranking covers 20 evaluated eval terminal from a public dataset that AI Battle re-checks daily.
Last updated: . Source: public benchmark datasets and provider pricing, aggregated daily; 20 entries shown.
| # | Entry | Creator | Value | Δ vs #1 |
|---|---|---|---|---|
| 1 | GPT-6 Astra (xhigh) | OpenAI | 0.60 | 0.0% |
| 2 | GPT-6 Astra (max) | OpenAI | 0.59 | -0.8% |
| 3 | Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback) | Anthropic | 0.55 | -7.6% |
| 4 | GPT-6 Astra (high) | OpenAI | 0.54 | -9.3% |
| 5 | Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback) | Anthropic | 0.52 | -12.7% |
| 6 | Claude Fable 5.1 (Adaptive Reasoning, High Effort, Default Fallback) | Anthropic | 0.52 | -12.7% |
| 7 | GPT-6 Astra (medium) | OpenAI | 0.49 | -16.9% |
| 8 | Claude Opus 5 (Adaptive Reasoning, Max Effort) | Anthropic | 0.49 | -17.8% |
| 9 | Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) | Anthropic | 0.42 | -28.8% |
| 10 | GLM-5.3 (max) | Z AI | 0.42 | -29.7% |
| 11 | GPT-5.6 Sol (max) | OpenAI | 0.40 | -33.1% |
| 12 | Qwen3.8 Max (0902) | Alibaba | 0.39 | -34.7% |
| 13 | GPT-5.6 Terra (max) | OpenAI | 0.35 | -40.7% |
| 14 | Muse Spark 1.3 (max) | Meta | 0.33 | -44.1% |
| 15 | GLM 5.3 Flash | Z AI | 0.33 | -44.9% |
| 16 | DeepSeek V4.1 Flash (Reasoning, Max Effort) | DeepSeek | 0.27 | -55.1% |
| 17 | Grok 4.6 (high) | SpaceXAI | 0.21 | -64.4% |
| 18 | Gemini 3.8 Flash (high) | 0.20 | -66.9% | |
| 19 | DeepSeek V4 Pro 0813 (Reasoning, Max Effort) | DeepSeek | 0.14 | -76.3% |
| 20 | Kimi K3 (max) | Kimi | 0.13 | -78.8% |
Across the 20 numeric entries in this window the median is 0.41 and the mean is 0.39, so the typical entry sits 0.19 below the leader. The window spans 0.13 to 0.60.
10 of 20 entries are at or above the median, and the top three hold 22% of the combined value — read it against the metric below, since a higher number is not "better" for every dataset.
9 creators appear in this window; the three most-represented hold 65% of the entries.
| Creator | Entries | Highest value | Top entry |
|---|---|---|---|
| OpenAI | 6 | 0.60 | GPT-6 Astra (xhigh) |
| Anthropic | 5 | 0.55 | Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback) |
| Z AI | 2 | 0.42 | GLM-5.3 (max) |
| DeepSeek | 2 | 0.27 | DeepSeek V4.1 Flash (Reasoning, Max Effort) |
| Alibaba | 1 | 0.39 | Qwen3.8 Max (0902) |
| Meta | 1 | 0.33 | Muse Spark 1.3 (max) |
| SpaceXAI | 1 | 0.21 | Grok 4.6 (high) |
| 1 | 0.20 | Gemini 3.8 Flash (high) | |
| Kimi | 1 | 0.13 | Kimi K3 (max) |
The same models are measured across multiple independent datasets. Aggregating the 20 entries above against every other AI Battle dataset gives this cross-section, computed by AI Battle and published nowhere upstream:
| Model | Datasets | Best placements elsewhere |
|---|---|---|
| GPT-6 Astra (xhigh) | 15 | #2/20 in Humanity's Last Exam: Output Tokens per Task, #3/20 in MMMU-Pro: Score, #3/20 in MMMU-Pro: Token Usage |
| GPT-6 Astra (max) | 54 | #1/20 in MMMU-Pro: Score, #2/20 in Artificial Analysis Intelligence Index, #2/20 in Artificial Analysis Intelligence Index by Open Weights / Proprietary |
| Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback) | 17 | #1/20 in AA-Briefcase Elo, #1/20 in GDPval-AA v2 Leaderboard, #2/20 in Humanity's Last Exam: Score |
| GPT-6 Astra (high) | 11 | #1/20 in AA-Omniscience Index: Score, #2/20 in MMMU-Pro: Score, #2/20 in MMMU-Pro: Token Usage |
| Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback) | 39 | #1/20 in Artificial Analysis Intelligence Index, #1/20 in Artificial Analysis Intelligence Index by Open Weights / Proprietary, #1/20 in AA-Briefcase Elo |
| Claude Fable 5.1 (Adaptive Reasoning, High Effort, Default Fallback) | 22 | #1/20 in SciCode: Output Tokens per Task, #2/20 in AA-Omniscience Index: Token Usage, #2/20 in AA-LCR v1.1: Output Tokens per Task |
| GPT-6 Astra (medium) | 12 | #1/20 in MMMU-Pro: Token Usage, #1/20 in AA-Omniscience Index: Token Usage, #2/20 in Terminal-Bench 4.0: Output Tokens per Task |
| Claude Opus 5 (Adaptive Reasoning, Max Effort) | 48 | #1/20 in GDPval-AA v2 Leaderboard, #2/20 in AA-Briefcase Elo, #2/20 in AA-Briefcase Elo |
| Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) | 31 | #2/20 in SciCode: Score, #3/20 in AA-Omniscience Index, #3/20 in AA-Omniscience Accuracy |
| GLM-5.3 (max) | 59 | #4/20 in AA-Briefcase Tool Calls Breakdown, Avg per Task, #4/20 in 𝜏³-Banking: Score, #5/20 in AA-Omniscience Hallucination Rate |
| GPT-5.6 Sol (max) | 58 | #3/20 in SciCode: Output Tokens per Task, #4/20 in 𝜏³-Banking: Output Tokens per Task, #5/20 in Artificial Analysis Intelligence Index |
| Qwen3.8 Max (0902) | 40 | #2/20 in Mean Turns per Task, #2/20 in AA-Briefcase Tool Calls Breakdown, Avg per Task, #2/20 in GDPval-AA v2: Average Turns per Task |
| GPT-5.6 Terra (max) | 52 | #7/20 in AA-Briefcase Output Tokens per Task, #8/20 in Humanity's Last Exam: Output Tokens per Task, #10/20 in Artificial Analysis Intelligence Index |
| Muse Spark 1.3 (max) | 62 | #3/20 in Output Speed, #3/20 in Output Speed, #3/20 in GDPval-AA v2 Leaderboard |
| GLM 5.3 Flash | 55 | #1/20 in Pricing: Cache Hit, Input, and Output, #1/20 in Pricing: Cache Hit, Input, and Output, #1/20 in AA-Omniscience Index: Cost Breakdown |
| DeepSeek V4.1 Flash (Reasoning, Max Effort) | 59 | #2/20 in End-to-End Response Time, #2/20 in Latency: Time To First Answer Token, #2/11 in Cost per Task |
| Grok 4.6 (high) | 64 | #2/20 in 𝜏³-Banking: Score, #4/20 in AA-Omniscience Index, #5/20 in Artificial Analysis Finance & Accounting Index |
| Gemini 3.8 Flash (high) | 63 | #1/11 in Speed, #1/11 in Speed, #2/20 in Output Speed |
| DeepSeek V4 Pro 0813 (Reasoning, Max Effort) | 57 | #2/13 in Model Size: Total and Active Parameters, #3/11 in Cost per Task, #3/11 in Cost per Task |
| Kimi K3 (max) | 62 | #1/20 in Context Window, #1/20 in AA-LCR v1.1: Score, #1/13 in Model Size: Total and Active Parameters |
This ranking carries the metric
Terminal-Bench 4.0 from the
Terminal-Bench 4.0: Score dataset in the tab::eval-terminal family. Units and the
direction of "better" are not declared by the source dataset, so AI Battle does not invent
them: values are shown exactly as published, and the ordering is the source dataset's own.
Rebuilt on from the
public dataset Terminal-Bench 4.0: Score (family tab::eval-terminal).
The source lists 20 entries and all 20 are shown here. AI Battle requests
the dataset's rows, keeps the ordering the source publishes, and computes only the statistics
on this page: median, mean, spread, the #1–#2 gap, the creator breakdown and the raw-data file.
Nothing on this page is generated text about the entries — the numbers come from the source.
Refresh cadence: every deploy re-pulls the dataset, so this page tracks the source's own update cycle. Corrections: see the methodology page.
Machine-readable copy of this table, with the statistics below and the source provenance: /data/rankings/eval-terminal-terminal-bench-4-0-score.json. Free to reuse with attribution to AI Battle and the source dataset.
GPT-6 Astra (xhigh) by OpenAI currently leads the Terminal-Bench 4.0: Score ranking with a score of 0.60, the highest value in this dataset across eval terminal. AI Battle aggregates the dataset from its public source and re-checks it daily.
The Terminal-Bench 4.0: Score ranking is rebuilt daily from the upstream public dataset. Scores are the source dataset's own published values, so positions track whatever that source last released — each ranking page names the dataset it comes from.
The top 3 are: 1. GPT-6 Astra (xhigh) (0.60), 2. GPT-6 Astra (max) (0.59), 3. Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback) (0.55).