Terminal-Bench 2.1 Extended — Mercor evaluation (Terminal-Bench 2.1 Extended (Mercor))

We show this table for reference; we do not rank on it.

Completing tasks in command-line environments. This table shows Mercor-run configurations for reference and is excluded from model rankings.

Evaluation results published by Mercor. Published evaluation aggregates only; task contents and dataset license grants are not included.

Pass@1 on Terminal-Bench 2.1 Extended (Mercor) — October 1, 2026 capture

We mirror the published pass@1 view for Terminal-Bench 2.1 Extended (Mercor). Opus 5.5 leads the public snapshot at 47.50%, followed by GPT-6.1 Sol (45.80%) and Grok 4.6 (45.50%). We do not use these results to rank models overall.

Opus 5.5

Anthropic

Mercor extended-set run · max reasoning

Published margin ±8.9 · Original set 83.5% ±7.7 · 297 reported samples

47.50%

GPT-6.1 Sol

OpenAI

Mercor extended-set run · max reasoning

Published margin ±8.9 · Original set 85.4% ±6.4 · 297 reported samples

45.80%

Grok 4.6

xAI

Mercor extended-set run · xhigh reasoning

Published margin ±8.8 · Original set 84.6% ±5.8 · 297 reported samples

45.50%

35 configurationsAgenticDated Mercor configurationsDisplay onlyUpdated October 1, 2026 capture

Pass@1 table (35 configurations)

Score

Opus 5.5AnthropicMercor extended-set run · max reasoningPublished margin ±8.9 · Original set 83.5% ±7.7 · 297 reported samples

47.50%

GPT-6.1 SolOpenAIMercor extended-set run · max reasoningPublished margin ±8.9 · Original set 85.4% ±6.4 · 297 reported samples

45.80%

Grok 4.6xAIMercor extended-set run · xhigh reasoningPublished margin ±8.8 · Original set 84.6% ±5.8 · 297 reported samples

45.50%

Sonnet 5.5AnthropicMercor extended-set run · max reasoningPublished margin ±9.1 · Original set 87.3% ±6.4 · 297 reported samples

45.50%

GPT-6 SolOpenAIMercor extended-set run · max reasoningPublished margin ±8.8 · Original set 85.4% ±6.7 · 296 reported samples

42.40%

Gemini 3.6 FlashGoogleMercor extended-set run · high reasoningPublished margin ±8.1 · Original set 74.5% ±7.9 · 297 reported samples

40.10%

GLM-5.3ZhipuMercor extended-set run · max reasoningPublished margin ±8.1 · Original set 83.5% ±6.2 · 297 reported samples

39.40%

DeepSeek-V4.1-FlashDeepSeekMercor extended-set run · max reasoningPublished margin ±8.2 · Original set 86.5% ±5.1 · 297 reported samples

38.40%

GPT-5.5OpenAIMercor extended-set run · xhigh reasoningPublished margin ±8.4 · Original set 83.1% ±6.6 · 297 reported samples

38.00%

GPT-5.6 SolOpenAIMercor extended-set run · xhigh reasoningPublished margin ±9.1 · Original set 84.6% ±6.2 · 297 reported samples

37.40%

GPT-5.6 TerraOpenAIMercor extended-set run · max reasoningPublished margin ±8.1 · Original set 80.9% ±6.9 · 297 reported samples

36.40%

GPT-5.6 LunaOpenAIMercor extended-set run · max reasoningPublished margin ±7.9 · Original set 76.4% ±7.5 · 297 reported samples

35.40%

Gemini 3.1 ProGoogleMercor extended-set run · high reasoningPublished margin ±8.1 · Original set 72.7% ±8.1 · 297 reported samples

35.00%

Opus 5AnthropicMercor extended-set run · high reasoningPublished margin ±8.6 · Original set 83.9% ±6.7 · 297 reported samples

34.00%

Grok 4.5xAIMercor extended-set run · high reasoningPublished margin ±7.9 · Original set 79% ±7.5 · 297 reported samples

33.30%

GPT-6 LunaOpenAIMercor extended-set run · max reasoningPublished margin ±8.2 · Original set 77.9% ±7.5 · 297 reported samples

33.30%

GPT-5.4OpenAIMercor extended-set run · xhigh reasoningPublished margin ±7.7 · Original set 77.2% ±7.9 · 297 reported samples

31.30%

Kimi K3KimiMercor extended-set run · max reasoningPublished margin ±7.1 · Original set 82% ±6.4 · 297 reported samples

31.00%

DeepSeek-V4-FlashDeepSeekMercor extended-set run · max reasoningPublished margin ±7.6 · Original set 72.3% ±8.4 · 297 reported samples

30.00%

Opus 4.8AnthropicMercor extended-set run · high reasoningPublished margin ±7.6 · Original set 82% ±6.7 · 297 reported samples

29.30%

Sonnet 5AnthropicMercor extended-set run · high reasoningPublished margin ±7.7 · Original set 82.4% ±7.1 · 294 reported samples

28.90%

Gemini 3.5 FlashGoogleMercor extended-set run · high reasoningPublished margin ±7.6 · Original set 73.8% ±7.5 · 297 reported samples

28.60%

Opus 4.7AnthropicMercor extended-set run · high reasoningPublished margin ±7.6 · Original set 76.4% ±7.9 · 297 reported samples

27.60%

Kimi K2.7 CodeKimiMercor extended-set run · high reasoningPublished margin ±6.6 · Original set 68.9% ±8.1 · 297 reported samples

23.60%

MiniMax-M3MiniMaxMercor extended-set run · high reasoningPublished margin ±6.2 · Original set 61.4% ±8.4 · 297 reported samples

22.20%

DeepSeek-V4-ProDeepSeekMercor extended-set run · max reasoningPublished margin ±6.2 · Original set 66.7% ±8.4 · 297 reported samples

21.20%

Fable 5.1AnthropicMercor extended-set run · max reasoningPublished margin ±7.2 · Original set 72.7% ±8.6 · 297 reported samples

19.50%

Qwen3.5AlibabaMercor extended-set runPublished margin ±6.2 · Original set 53.9% ±9.7 · 297 reported samples

17.80%

Sonnet 4.6AnthropicMercor extended-set run · high reasoningPublished margin ±6.2 · Original set 62.9% ±9.4 · 297 reported samples

17.50%

Fable 5AnthropicMercor extended-set run · max reasoningPublished margin ±6.7 · Original set not reported · 297 reported samples

16.80%

DeepSeek-V3.2DeepSeekMercor extended-set runPublished margin ±5.1 · Original set 44.6% ±9 · 297 reported samples

15.20%

Nemotron 3 UltraNVIDIAMercor extended-set run · high reasoningPublished margin ±5.1 · Original set 49.8% ±9 · 297 reported samples

11.80%

Grok 4.7xAIMercor extended-set run · xhigh reasoningPublished margin ±4.2 · Original set not reported · 297 reported samples

11.80%

Kimi K2 (Thinking)KimiMercor extended-set run · high reasoningPublished margin ±4.5 · Original set 39% ±8.6 · 297 reported samples

11.10%

MiniMax-M2.7MiniMaxMercor extended-set run · high reasoningPublished margin ±4.7 · Original set 54.3% ±9 · 297 reported samples

10.10%

How to read this Mercor evaluation

We captured Mercor's Terminal-Bench 2.1 Extended results on October 1, 2026. The table preserves 35 published configurations, their source identifiers, reasoning settings, and disclosed pass@1 values. These are results from Mercor's evaluation setup; we did not rerun them.

The headline uses the 99 Mercor-built tasks. The paired original-set values refer to the separately listed 89-task public benchmark. Missing baselines remain unreported. We keep the extension and original set separate from their upstream benchmark-owner and provider-run tables.

The headline is the source page's selected Pass@1 metric. The task set, harness, tools, and grader belong to this Mercor run and may differ from another evaluator's results.

The row labels retain published error margins and sample counts when supplied. Missing margins and counts remain unreported; a sample count is not used to reconstruct a task denominator. Except where the dedicated source defines an interval, the confidence level is unreported. Equal displayed scores and overlapping margins do not establish statistical ties. Capture time does not establish individual evaluation dates.

The general methodology lists programmatic unit tests (all must pass) and a judge of None. Its run-count label is k = 3. Those notes describe the original public benchmark; extension-specific changes are not established by that reference.

Snapshot

Pass@135 configurationsMercor task extensionDisplay only

Mercor Terminal-Bench 2.1 Extended results Mercor APEX benchmark index Mercor methodology Paired Mercor original-set results Source paper Upstream evaluation code Upstream dataset

The published Terminal-Bench 2.1 Extended (Mercor) snapshot places Opus 5.5 first at 47.50%. The third row is 2.00 points behind. The broader top-10 range is 10.10 points, so the table still separates the published systems.

35 configurations are shown for Terminal-Bench 2.1 Extended (Mercor). The benchmark falls in the Agentic category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. Terminal-Bench 2.1 Extended (Mercor) is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About Terminal-Bench 2.1 Extended (Mercor)

Tasks

99 Mercor tasks; 89 tasks in the separately reported original set

Format

Pass@1

Difficulty

Source-specific evaluation

The selected headline is Pass@1. Source model identifiers, effort settings, published margins, and reported sample counts remain attached to the source configuration. Public-set results and Mercor task extensions occupy separate tables.

Terminal-Bench 2.1 Extended evaluation results Public benchmark source

Freshness and provenance

Version

Terminal-Bench 2.1 Extended; Mercor Pass@1

Refresh cadence

Reviewed dated source capture

Staleness state

Dated Mercor configurations

Question availability

Upstream task access and terms; only aggregate results mirrored

Dated Mercor configurationsDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does Terminal-Bench 2.1 Extended (Mercor) measure?

Completing tasks in command-line environments. This table shows Mercor-run configurations for reference and is excluded from model rankings.

Which model leads the published Terminal-Bench 2.1 Extended (Mercor) snapshot?

Opus 5.5 currently leads the published Terminal-Bench 2.1 Extended (Mercor) snapshot with 47.50% pass@1. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on Terminal-Bench 2.1 Extended (Mercor)?

The October 1, 2026 capture snapshot contains 35 source configurations.

Last updated: October 1, 2026 capture · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.