AI Battle > Rankings > Terminal-Bench v4.0: Output Tokens per Task
gpt-oss-120b (high) by OpenAI leads the Terminal-Bench v4.0: Output Tokens per Task ranking with a score of 2105. 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-oss-120b (high) | OpenAI | 2105 | 0.0% |
| 2 | GPT-6 Astra (medium) | OpenAI | 13648 | 548.3% |
| 3 | GPT-6 Astra (high) | OpenAI | 15322 | 627.8% |
| 4 | GPT-6 Astra (xhigh) | OpenAI | 17246 | 719.2% |
| 5 | Muse Glimmer (high) | Meta | 17743 | 742.7% |
| 6 | Nemotron 3 Ultra 550B A55B (Reasoning) | NVIDIA | 34172 | 1523.1% |
| 7 | Gemini 3.5 Flash-Lite | 26783 | 1172.1% | |
| 8 | Mistral Medium 3.5 | Mistral | 38059 | 1707.7% |
| 9 | GPT-6 Astra (max) | OpenAI | 23336 | 1008.4% |
| 10 | GPT-5.6 Sol (max) | OpenAI | 37485 | 1680.4% |
| 11 | Inkling (xhigh) | Thinking Machines | 25951 | 1132.6% |
| 12 | Grok 4.6 (high) | SpaceXAI | 36898 | 1652.6% |
| 13 | Claude Fable 5.1 (Adaptive Reasoning, High Effort, Default Fallback) | Anthropic | 43743 | 1977.7% |
| 14 | Kimi K3 (max) | Kimi | 30360 | 1342.0% |
| 15 | GPT-5.6 Terra (max) | OpenAI | 48466 | 2202.0% |
| 16 | GPT-5.6 Luna (max) | OpenAI | 41119 | 1853.1% |
| 17 | Qwen3.8 27B (xhigh) | Alibaba | 41701 | 1880.7% |
| 18 | Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback) | Anthropic | 55102 | 2517.2% |
| 19 | DeepSeek V4 Pro 0813 (Reasoning, Max Effort) | DeepSeek | 69376 | 3195.2% |
| 20 | MiniMax-M3 | MiniMax | 98158 | 4562.3% |
Across the 20 numeric entries in this window the median is 35535 and the mean is 35839, so the typical entry sits 33429 above the leader. The window spans 2105 to 98158.
10 of 20 entries are at or above the median, and the top three hold 31% of the combined value — read it against the metric below, since a higher number is not "better" for every dataset.
12 creators appear in this window; the three most-represented hold 55% of the entries.
| Creator | Entries | Highest value | Top entry |
|---|---|---|---|
| OpenAI | 8 | 48466 | GPT-5.6 Terra (max) |
| Anthropic | 2 | 55102 | Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback) |
| MiniMax | 1 | 98158 | MiniMax-M3 |
| DeepSeek | 1 | 69376 | DeepSeek V4 Pro 0813 (Reasoning, Max Effort) |
| Alibaba | 1 | 41701 | Qwen3.8 27B (xhigh) |
| Mistral | 1 | 38059 | Mistral Medium 3.5 |
| SpaceXAI | 1 | 36898 | Grok 4.6 (high) |
| NVIDIA | 1 | 34172 | Nemotron 3 Ultra 550B A55B (Reasoning) |
| Kimi | 1 | 30360 | Kimi K3 (max) |
| 1 | 26783 | Gemini 3.5 Flash-Lite | |
| Thinking Machines | 1 | 25951 | Inkling (xhigh) |
| Meta | 1 | 17743 | Muse Glimmer (high) |
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-oss-120b (high) | 39 | #1/20 in Cost to Run Artificial Analysis Intelligence Index, #1/20 in Cost to Run Artificial Analysis Intelligence Index, #1/20 in Humanity's Last Exam: Cost per Task |
| GPT-6 Astra (medium) | 12 | #1/20 in AA-Omniscience Index: Token Usage, #1/19 in MMMU-Pro: Token Usage, #6/19 in MMMU-Pro: Score |
| GPT-6 Astra (high) | 12 | #1/20 in AA-Omniscience Index: Score, #2/19 in MMMU-Pro: Token Usage, #2/19 in MMMU-Pro: Score |
| GPT-6 Astra (xhigh) | 15 | #1/20 in Terminal-Bench v4.0: Score, #2/20 in Humanity's Last Exam: Output Tokens per Task, #3/20 in AA-Omniscience Index: Score |
| Muse Glimmer (high) | 43 | #1/20 in Output Tokens per Intelligence Index Task, #1/20 in Cost per Intelligence Index Task, #1/20 in Cost per Intelligence Index Task |
| Nemotron 3 Ultra 550B A55B (Reasoning) | 46 | #1/12 in Artificial Analysis Openness Index: Score, #1/12 in Artificial Analysis Openness Index: Components, #2/20 in 𝜏³-Banking: Time per Task |
| Gemini 3.5 Flash-Lite | 44 | #1/20 in End-to-End Response Time, #1/20 in Output Speed, #1/20 in Time per Intelligence Index Task |
| Mistral Medium 3.5 | 41 | #2/20 in GDPval-AA v2: Output Tokens per Task, #3/20 in Output Tokens per Intelligence Index Task, #3/20 in Output Tokens per Intelligence Index Task |
| GPT-6 Astra (max) | 51 | #1/19 in MMMU-Pro: Score, #2/20 in AA-Omniscience Index, #2/20 in Artificial Analysis Intelligence Index by Open Weights / Proprietary |
| GPT-5.6 Sol (max) | 56 | #3/20 in SciCode: Output Tokens per Task, #4/20 in 𝜏³-Banking: Output Tokens per Task, #5/20 in AA-LCR v1.1: Output Tokens per Task |
| Inkling (xhigh) | 52 | #3/20 in Context Window, #3/20 in 𝜏³-Banking: Output Tokens per Task, #5/20 in AA-Briefcase Output Tokens per Task |
| Grok 4.6 (high) | 61 | #2/20 in 𝜏³-Banking: Score, #5/20 in AA-Briefcase Elo, #5/20 in AA-Omniscience Index |
| 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 |
| Kimi K3 (max) | 60 | #1/20 in Context Window, #1/20 in AA-LCR v1.1: Score, #1/14 in Model Size: Total and Active Parameters |
| 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, #9/20 in Output Speed |
| GPT-5.6 Luna (max) | 58 | #1/11 in Cost per Task, #1/11 in Cost per Task, #3/20 in Pricing: Cache Hit, Input, and Output |
| Qwen3.8 27B (xhigh) | 43 | #6/20 in AA-Omniscience Hallucination Rate, #6/20 in 𝜏³-Banking: Score, #8/20 in AA-LCR v1.1: Cost per Task |
| Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback) | 20 | #2/20 in Humanity's Last Exam: Score, #2/20 in AA-Omniscience Accuracy, #2/20 in AA-Briefcase Rubric Pass Rate by File Type (Normalized) |
| DeepSeek V4 Pro 0813 (Reasoning, Max Effort) | 59 | #3/14 in Model Size: Total and Active Parameters, #3/12 in Artificial Analysis Openness Index: Score, #3/11 in Cost per Task |
| MiniMax-M3 | 57 | #1/20 in AA-Omniscience Hallucination Rate, #1/20 in 𝜏³-Banking: Output Tokens per Task, #4/20 in Pricing: Cache Hit, Input, and Output |
This ranking carries the metrics
answer, reasoning from the
Terminal-Bench v4.0: Output Tokens per Task 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 v4.0: Output Tokens per Task (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-v4-0-output-tokens-per-task.json. Free to reuse with attribution to AI Battle and the source dataset.
gpt-oss-120b (high) by OpenAI currently leads the Terminal-Bench v4.0: Output Tokens per Task ranking with a score of 2105, 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 v4.0: Output Tokens per Task 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-oss-120b (high) (2105), 2. GPT-6 Astra (medium) (13648), 3. GPT-6 Astra (high) (15322).