CharXiv — Mercor evaluation (CharXiv (Mercor))

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

Answering descriptive and reasoning questions about research-paper charts. 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 CharXiv (Mercor) — October 15, 2026 capture

We mirror the published pass@1 view for CharXiv (Mercor). Opus 5.5 leads the public snapshot at 96.40%, followed by GPT-6 Astra (96.10%) and Sonnet 5.5 (96.10%). We do not use these results to rank models overall.

Opus 5.5

Anthropic

Single-shot multimodal · max reasoning

Published margin ±0.5 · 15,000 reported samples

96.40%

GPT-6 Astra

OpenAI

Single-shot multimodal · max reasoning

Published margin ±0.5 · 15,000 reported samples

96.10%

Sonnet 5.5

Anthropic

Single-shot multimodal · max reasoning

Published margin ±0.5 · 14,999 reported samples

96.10%

40 configurationsMultimodal & GroundedDated Mercor configurationsDisplay onlyUpdated October 15, 2026 capture

Pass@1 table (40 configurations)

Pass@1 results for CharXiv (Mercor)
Rank	Model / configuration	Score	Parameters (B)	Open / closed
1	Opus 5.5AnthropicSingle-shot multimodal · max reasoningPublished margin ±0.5 · 15,000 reported samples	96.40%	Not reported	Unknown
2	GPT-6 AstraOpenAISingle-shot multimodal · max reasoningPublished margin ±0.5 · 15,000 reported samples	96.10%	Not reported	Unknown
3	Sonnet 5.5AnthropicSingle-shot multimodal · max reasoningPublished margin ±0.5 · 14,999 reported samples	96.10%	Not reported	Unknown
4	GPT-6.1 SolOpenAISingle-shot multimodal · max reasoningPublished margin ±0.5 · 14,986 reported samples	96.00%	Not reported	Unknown
5	GPT-6 AstraOpenAISingle-shot multimodal · xhigh reasoningPublished margin ±0.5 · 15,000 reported samples	95.80%	Not reported	Unknown
6	Gemini 3.8 FlashGoogleSingle-shot multimodal · high reasoningPublished margin ±0.5 · 15,000 reported samples	95.40%	Not reported	Unknown
7	Gemini 3.7 FlashGoogleSingle-shot multimodal · high reasoningPublished margin ±0.5 · 15,000 reported samples	95.30%	Not reported	Unknown
8	Muse Spark 1.2MetaSingle-shot multimodal · xhigh reasoningPublished margin ±0.5 · 15,000 reported samples	95.30%	Not reported	Unknown
9	Muse Spark 1.1MetaSingle-shot multimodal · xhigh reasoningPublished margin ±0.5 · 15,000 reported samples	95.20%	Not reported	Unknown
10	Gemini 3.6 FlashGoogleSingle-shot multimodal · high reasoningPublished margin ±0.6 · 15,000 reported samples	95.10%	Not reported	Unknown
11	Fable 5.1AnthropicSingle-shot multimodal · max reasoningPublished margin ±0.6 · 15,000 reported samples	95.10%	Not reported	Unknown
12	GPT-6 SolOpenAISingle-shot multimodal · max reasoningPublished margin ±0.6 · 15,000 reported samples	95.00%	Not reported	Unknown
13	Opus 5AnthropicSingle-shot multimodal · max reasoningPublished margin ±0.6 · 14,998 reported samples	94.60%	Not reported	Unknown
14	GPT-5.5OpenAISingle-shot multimodal · xhigh reasoningPublished margin ±0.6 · 14,998 reported samples	94.60%	Not reported	Unknown
15	Gemini 3.1 ProGoogleSingle-shot multimodal · high reasoningPublished margin ±0.6 · 14,998 reported samples	94.60%	Not reported	Unknown
16	Gemini 3.5 FlashGoogleSingle-shot multimodal · high reasoningPublished margin ±0.6 · 14,999 reported samples	94.30%	Not reported	Unknown
17	Qwen3.8-MaxAlibabaSingle-shot multimodal · xhigh reasoningPublished margin ±0.6 · 15,000 reported samples	94.10%	Not reported	Unknown
18	GPT-5.6 TerraOpenAISingle-shot multimodal · max reasoningPublished margin ±0.6 · 14,998 reported samples	94.10%	Not reported	Unknown
19	Fable 5.1AnthropicSingle-shot multimodal · high reasoningPublished margin ±0.6 · 15,000 reported samples	94.10%	Not reported	Unknown
20	Haiku 5.5AnthropicSingle-shot multimodal · max reasoningPublished margin ±0.6 · 15,000 reported samples	94.10%	Not reported	Unknown
21	Kimi K3KimiSingle-shot multimodal · max reasoningPublished margin ±0.6 · 15,000 reported samples	93.90%	Not reported	Unknown
22	Muse Spark 1.3MetaSingle-shot multimodal · max reasoningPublished margin ±0.6 · 14,993 reported samples	93.90%	Not reported	Unknown
23	GPT-5.4OpenAISingle-shot multimodal · xhigh reasoningPublished margin ±0.6 · 14,998 reported samples	93.40%	Not reported	Unknown
24	Opus 4.8AnthropicSingle-shot multimodal · max reasoningPublished margin ±0.6 · 14,991 reported samples	93.20%	Not reported	Unknown
25	GPT-5.6 LunaOpenAISingle-shot multimodal · max reasoningPublished margin ±0.6 · 15,000 reported samples	92.90%	Not reported	Unknown
26	Grok 4.6xAISingle-shot multimodal · xhigh reasoningPublished margin ±0.6 · 15,000 reported samples	92.90%	Not reported	Unknown
27	Grok 4.5xAISingle-shot multimodal · high reasoningPublished margin ±0.7 · 14,999 reported samples	92.80%	Not reported	Unknown
28	Opus 4.7AnthropicSingle-shot multimodal · max reasoningPublished margin ±0.7 · 14,999 reported samples	92.60%	Not reported	Unknown
29	GLM-5.3-FlashZhipuSingle-shot multimodal · max reasoningPublished margin ±0.7 · 14,998 reported samples	92.50%	Not reported	Unknown
30	GPT-5.6 Sol (Pro)OpenAISingle-shot multimodal · max reasoningPublished margin ±0.6 · 14,998 reported samples	92.10%	Not reported	Unknown
31	Grok 4.7xAISingle-shot multimodal · xhigh reasoningPublished margin ±0.7 · 14,648 reported samples	92.00%	Not reported	Unknown
32	Qwen3.5AlibabaSingle-shot multimodalPublished margin ±0.7 · 14,995 reported samples	91.60%	Not reported	Unknown
33	GPT-6 LunaOpenAISingle-shot multimodal · max reasoningPublished margin ±0.7 · 15,000 reported samples	90.90%	Not reported	Unknown
34	Sonnet 4.6AnthropicSingle-shot multimodal · high reasoningPublished margin ±0.8 · 14,999 reported samples	90.50%	Not reported	Unknown
35	DeepSeek-V4.1-FlashDeepSeekSingle-shot multimodal · max reasoningPublished margin ±0.7 · 15,000 reported samples	90.10%	Not reported	Unknown
36	Fable 5AnthropicSingle-shot multimodal · max reasoningPublished margin ±0.8 · 20,000 reported samples	89.10%	Not reported	Unknown
37	InklingThinking MachinesSingle-shot multimodal · high reasoningPublished margin ±0.8 · 15,000 reported samples	88.90%	Not reported	Unknown
38	Sonnet 5AnthropicSingle-shot multimodal · max reasoningPublished margin ±0.7 · 14,997 reported samples	88.80%	Not reported	Unknown
39	MiniMax-M3MiniMaxSingle-shot multimodal · high reasoningPublished margin ±0.8 · 15,000 reported samples	85.90%	Not reported	Unknown
40	GLM-5.3ZhipuSingle-shot multimodal · max reasoningPublished margin ±1.1 · 14,965 reported samples	63.00%	Not reported	Unknown

How to read this Mercor evaluation

We captured Mercor's CharXiv results on October 15, 2026. The table preserves 40 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.

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 llm-as-a-judge (lmaaj) and a judge of Gemini 3.5 Flash. Its run-count label is k = 3. The linked result page takes precedence where its setup differs, as noted above.

Snapshot

Pass@140 configurationsMercor evaluationDisplay only

Mercor CharXiv results Mercor APEX benchmark index Mercor methodology Source paper Upstream evaluation code Upstream dataset

The published CharXiv (Mercor) snapshot places Opus 5.5 first at 96.40%. The third row is 0.30 points behind. The broader top-10 range is 1.30 points, so many of the published results sit in a relatively narrow band.

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

About CharXiv (Mercor)

Tasks

5,000 source-reported 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.

CharXiv evaluation results Public benchmark source

Freshness and provenance

Version

CharXiv; 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 CharXiv (Mercor) measure?

Answering descriptive and reasoning questions about research-paper charts. This table shows Mercor-run configurations for reference and is excluded from model rankings.

Which model leads the published CharXiv (Mercor) snapshot?

Opus 5.5 currently leads the published CharXiv (Mercor) snapshot with 96.40% pass@1. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on CharXiv (Mercor)?

The October 15, 2026 capture snapshot contains 40 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.