APEX-SWE (Terminus-2) — Mercor evaluation (APEX-SWE (Terminus-2) (Mercor))

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

Software integration and production-failure diagnosis with the Terminus-2 harness. 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 APEX-SWE (Terminus-2) (Mercor) — October 15, 2026 capture

We mirror the published pass@1 view for APEX-SWE (Terminus-2) (Mercor). Opus 5.5 leads the public snapshot at 67.60%, followed by Sonnet 5.5 (66.40%) and Opus 5 (63.70%). We do not use these results to rank models overall.

Opus 5.5

Anthropic

Terminus-2 · max reasoning

Published margin ±5.9 · 800 reported samples

67.60%

Sonnet 5.5

Anthropic

Terminus-2 · max reasoning

Published margin ±6.1 · 800 reported samples

66.40%

Opus 5

Anthropic

Terminus-2 · max reasoning

Published margin ±6.4 · 800 reported samples

63.70%

57 configurationsCodingDated Mercor configurationsDisplay onlyUpdated October 15, 2026 capture

Pass@1 table (57 configurations)

Pass@1 results for APEX-SWE (Terminus-2) (Mercor)
Rank	Model / configuration	Score	Parameters (B)	Open / closed
1	Opus 5.5AnthropicTerminus-2 · max reasoningPublished margin ±5.9 · 800 reported samples	67.60%	Not reported	Unknown
2	Sonnet 5.5AnthropicTerminus-2 · max reasoningPublished margin ±6.1 · 800 reported samples	66.40%	Not reported	Unknown
3	Opus 5AnthropicTerminus-2 · max reasoningPublished margin ±6.4 · 800 reported samples	63.70%	Not reported	Unknown
4	Fable 5.1AnthropicTerminus-2 · max reasoningPublished margin ±6.3 · 776 reported samples	63.60%	Not reported	Unknown
5	Fable 5AnthropicTerminus-2 · max reasoningPublished margin ±6.4 · 799 reported samples	58.80%	Not reported	Unknown
6	Grok 4.6xAITerminus-2 · xhigh reasoningPublished margin ±6.2 · 1,600 reported samples	57.70%	Not reported	Unknown
7	Grok 4.6xAITerminus-2 · high reasoningPublished margin ±6.2 · 799 reported samples	56.40%	Not reported	Unknown
8	Opus 5.5AnthropicTerminus-2 · medium reasoningPublished margin ±6.3 · 801 reported samples	55.90%	Not reported	Unknown
9	Haiku 5.5AnthropicTerminus-2 · max reasoningPublished margin ±6.5 · 800 reported samples	55.10%	Not reported	Unknown
10	MiMo V2.6 ProXiaomiTerminus-2 · auto reasoningPublished margin ±6.2 · 793 reported samples	54.20%	Not reported	Unknown
11	Grok 4.5xAITerminus-2 · high reasoningPublished margin ±6.4 · 800 reported samples	53.60%	Not reported	Unknown
12	Grok 4.7xAITerminus-2 · xhigh reasoningPublished margin ±6.3 · 800 reported samples	53.40%	Not reported	Unknown
13	Qwen3.8-MaxAlibabaTerminus-2 · xhigh reasoningPublished margin ±6.4 · 799 reported samples	52.30%	Not reported	Unknown
14	DeepSeek-V4.1-FlashDeepSeekTerminus-2 · max reasoningPublished margin ±6.3 · 800 reported samples	51.90%	Not reported	Unknown
15	Qwen 3.8 27BAlibabaTerminus-2 · xhigh reasoningPublished margin ±6.2 · 800 reported samples	50.70%	Not reported	Unknown
16	Gemini 4 ArgonGoogleTerminus-2 · high reasoningPublished margin ±6.6 · 800 reported samples	49.20%	Not reported	Unknown
17	GPT-6 AstraOpenAITerminus-2 · max reasoningPublished margin ±6.3 · 1,600 reported samples	48.60%	Not reported	Unknown
18	Kimi K3KimiTerminus-2 · max reasoningPublished margin ±6.2 · 800 reported samples	48.00%	Not reported	Unknown
19	GPT-5.6 SolOpenAITerminus-2 · max reasoningPublished margin ±6.4 · 800 reported samples	48.00%	Not reported	Unknown
20	Opus 4.7AnthropicTerminus-2 · max reasoningPublished margin ±6.5 · 800 reported samples	47.60%	Not reported	Unknown
21	GPT-6.1 SolOpenAITerminus-2 · max reasoningPublished margin ±6.4 · 800 reported samples	46.90%	Not reported	Unknown
22	Sonnet 5AnthropicTerminus-2 · max reasoningPublished margin ±6.4 · 800 reported samples	46.40%	Not reported	Unknown
23	GLM-5.3-FlashZhipuTerminus-2 · max reasoningPublished margin ±6.1 · 800 reported samples	46.10%	Not reported	Unknown
24	DeepSeek-V4-Pro-0813DeepSeekTerminus-2 · max reasoningPublished margin ±6.1 · 1,570 reported samples	45.90%	Not reported	Unknown
25	GPT-5.6 SolOpenAITerminus-2 · xhigh reasoningPublished margin ±6.2 · 800 reported samples	45.80%	Not reported	Unknown
26	GLM-5.3ZhipuTerminus-2 · max reasoningPublished margin ±5.9 · 800 reported samples	45.60%	Not reported	Unknown
27	GPT-6 SolOpenAITerminus-2 · max reasoningPublished margin ±6.3 · 800 reported samples	45.00%	Not reported	Unknown
28	Sonnet 5.5AnthropicTerminus-2 · medium reasoningPublished margin ±6.1 · 800 reported samples	44.00%	Not reported	Unknown
29	Opus 4.8AnthropicTerminus-2 · max reasoningPublished margin ±6.1 · 800 reported samples	43.90%	Not reported	Unknown
30	Gemini 3.7 FlashGoogleTerminus-2 · high reasoningPublished margin ±6.4 · 800 reported samples	42.00%	Not reported	Unknown
31	Opus 4.6AnthropicTerminus-2 · max reasoningPublished margin ±6.2 · 795 reported samples	41.80%	Not reported	Unknown
32	Muse Spark 1.2MetaTerminus-2 · xhigh reasoningPublished margin ±6.2 · 800 reported samples	40.50%	Not reported	Unknown
33	Kimi K2.7 CodeKimiTerminus-2 · high reasoningPublished margin ±5.9 · 800 reported samples	40.30%	Not reported	Unknown
34	Gemini 3.6 FlashGoogleTerminus-2 · high reasoningPublished margin ±6.2 · 800 reported samples	39.40%	Not reported	Unknown
35	Muse Spark 1.3MetaTerminus-2 · max reasoningPublished margin ±6.4 · 1,592 reported samples	39.10%	Not reported	Unknown
36	Muse Spark 1.1MetaTerminus-2 · xhigh reasoningPublished margin ±6.1 · 656 reported samples	38.90%	Not reported	Unknown
37	GPT-6 LunaOpenAITerminus-2 · max reasoningPublished margin ±6.1 · 800 reported samples	38.80%	Not reported	Unknown
38	GLM-5.2ZhipuTerminus-2 · max reasoningPublished margin ±6.1 · 800 reported samples	37.40%	Not reported	Unknown
39	GPT-5.5OpenAITerminus-2 · xhigh reasoningPublished margin ±5.8 · 1,600 reported samples	36.60%	Not reported	Unknown
40	Gemini 3.8 FlashGoogleTerminus-2 · high reasoningPublished margin ±5.9 · 1,595 reported samples	36.40%	Not reported	Unknown
41	Muse Spark 1.3MetaTerminus-2 · xhigh reasoningPublished margin ±6.4 · 796 reported samples	36.40%	Not reported	Unknown
42	Gemini 3.5 FlashGoogleTerminus-2 · high reasoningPublished margin ±6 · 800 reported samples	36.10%	Not reported	Unknown
43	MiniMax-M3MiniMaxTerminus-2 · high reasoningPublished margin ±5.6 · 1,600 reported samples	34.60%	Not reported	Unknown
44	GLM-5.1ZhipuTerminus-2Published margin ±5.7 · 800 reported samples	34.50%	Not reported	Unknown
45	Sonnet 4.6AnthropicTerminus-2 · high reasoningPublished margin ±6.1 · 800 reported samples	34.50%	Not reported	Unknown
46	Gemini 3.1 ProGoogleTerminus-2 · high reasoningPublished margin ±5.8 · 800 reported samples	33.90%	Not reported	Unknown
47	DeepSeek-V4-ProDeepSeekTerminus-2 · max reasoningPublished margin ±5.7 · 800 reported samples	33.80%	Not reported	Unknown
48	GPT-5.6 LunaOpenAITerminus-2 · max reasoningPublished margin ±5.6 · 800 reported samples	32.10%	Not reported	Unknown
49	Gemini 3.5 Flash LiteGoogleTerminus-2 · high reasoningPublished margin ±5.6 · 800 reported samples	31.20%	Not reported	Unknown
50	GPT-5.6 TerraOpenAITerminus-2 · max reasoningPublished margin ±5.6 · 800 reported samples	31.10%	Not reported	Unknown
51	Qwen3.5AlibabaTerminus-2Published margin ±5.1 · 2,399 reported samples	27.90%	Not reported	Unknown
52	InklingThinking MachinesTerminus-2 · high reasoningPublished margin ±5.2 · 800 reported samples	26.10%	Not reported	Unknown
53	DeepSeek-V3.2DeepSeekTerminus-2Published margin ±5.1 · 800 reported samples	25.60%	Not reported	Unknown
54	MiniMax-M2.7MiniMaxTerminus-2 · high reasoningPublished margin ±4.6 · 800 reported samples	21.50%	Not reported	Unknown
55	Kimi K2 (Thinking)KimiTerminus-2 · high reasoningPublished margin ±4.2 · 800 reported samples	17.40%	Not reported	Unknown
56	Nemotron 3 UltraNVIDIATerminus-2 · high reasoningPublished margin ±3.6 · 1,579 reported samples	15.20%	Not reported	Unknown
57	GPT-OSS-120BOpenAITerminus-2 · high reasoningPublished margin ±1.6 · 800 reported samples	3.30%	Not reported	Unknown

How to read this Mercor evaluation

We captured Mercor's APEX-SWE (Terminus-2) results on October 15, 2026. The table preserves 57 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.

This dated source table is display only and does not enter overall or category rankings. Source configurations retain their external identity, including Pro descriptors and different effort settings.

These are the disclosed Terminus-2 Pass@1 results across integration and observability work. Inspect is a separate harness profile. Configurations without a score for the selected harness stay absent, rather than receiving zero.

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 expert-authored rubrics plus repository test suites and a judge of LLM judge + programmatic tests. Its run-count label is k = 3. The linked result page takes precedence where its setup differs, as noted above.

Snapshot

Pass@157 configurationsMercor evaluationDisplay only

Mercor APEX-SWE (Terminus-2) results Mercor APEX benchmark index Mercor methodology Source paper Upstream evaluation code Upstream dataset Mercor evaluation notes

The published APEX-SWE (Terminus-2) (Mercor) snapshot places Opus 5.5 first at 67.60%. The third row is 3.90 points behind. The broader top-10 range is 13.40 points, so the table still separates the published systems.

57 configurations are shown for APEX-SWE (Terminus-2) (Mercor). The benchmark falls in the Coding category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. APEX-SWE (Terminus-2) (Mercor) is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About APEX-SWE (Terminus-2) (Mercor)

Tasks

Integration and observability engineering tasks

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.

APEX-SWE (Terminus-2) evaluation results Public benchmark source

Freshness and provenance

Version

APEX-SWE (Terminus-2); Mercor Pass@1; terminus-2

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 APEX-SWE (Terminus-2) (Mercor) measure?

Software integration and production-failure diagnosis with the Terminus-2 harness. This table shows Mercor-run configurations for reference and is excluded from model rankings.

Which model leads the published APEX-SWE (Terminus-2) (Mercor) snapshot?

Opus 5.5 currently leads the published APEX-SWE (Terminus-2) (Mercor) snapshot with 67.60% pass@1. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on APEX-SWE (Terminus-2) (Mercor)?

The October 15, 2026 capture snapshot contains 57 source configurations.

Last updated: October 15, 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 5,600+ readers.

One email each week. Unsubscribe anytime.