Market intelligence · Text

Benchmark Explorer

Compare model adoption, open-source momentum, tasks, benchmarks, and runtime performance with visible source coverage, privacy controls, and versioned methodology.

OpenRouterSource snapshot · 2026-07-26 00:00 UTC

Arc.AIAggregate snapshot · 2026-07-26 00:00 UTC

CombinedAdoption index · 60% OR / 40% Arc

Usage trend

Top Models

Weekly adoption trend with source-specific coverage and a versioned Combined index.

Apr 20

Apr 27

May 4

May 11

May 18

May 25

Jun 1

Jun 8

Jun 15

Jun 22

Jun 29

Jul 6

Jul 13

Jul 20

Combined displays a versioned adoption index. It does not add OpenRouter and Arc.AI token totals.

Popularity

LLM Leaderboard

Compare ranked model variants with visible source contribution, coverage, and privacy status. Period-over-period change remains unavailable until matched prior snapshots exist.

RankModelOpennessSource contributionBlended index

1Commercial open weights 10.5T tokensComplete for selected window · 2026-07-26 00:00 UTC 210B tokensComplete for selected window · 2026-07-26 00:00 UTC69.5 index

2Commercial open weights 3.29T tokensComplete for selected window · 2026-07-26 00:00 UTC 880B tokensComplete for selected window · 2026-07-26 00:00 UTC58.8 index

3OSI-aligned candidate 6.37T tokensComplete for selected window · 2026-07-26 00:00 UTC 420B tokensComplete for selected window · 2026-07-26 00:00 UTC55.5 index

4Proprietary 1.44T tokensComplete for selected window · 2026-07-26 00:00 UTC 620B tokensComplete for selected window · 2026-07-26 00:00 UTC36.4 index

5OSI-aligned candidate 3.17T tokensComplete for selected window · 2026-07-26 00:00 UTC 280B tokensComplete for selected window · 2026-07-26 00:00 UTC30.8 index

6Restricted open weights 3.94T tokensComplete for selected window · 2026-07-26 00:00 UTC 130B tokensComplete for selected window · 2026-07-26 00:00 UTC28.4 index

7Commercial open weights 2.05T tokensComplete for selected window · 2026-07-26 00:00 UTC 360B tokensComplete for selected window · 2026-07-26 00:00 UTC28.1 index

8Commercial open weights 1.92T tokensComplete for selected window · 2026-07-26 00:00 UTC 240B tokensComplete for selected window · 2026-07-26 00:00 UTC21.9 index

Combined is a transparent index—not a raw token sum. OpenRouter and Arc.AI known contributions and suppression states remain visible separately to avoid false precision and upstream double counting. Day and 30-day values use the same versioned source window and normalization rules.

Arc.AI exclusive view

Open Source Ranking

An OSI-aligned review queue, separate from broader commercial or restricted open-weight models.

OSI-aligned review4Exact variants · evidence record attached

Commercial open weights4Visible in Open Models, not strict Open Source

License evidenceRequiredSPDX, evidence URL, verifier, and review date

Open weights do not automatically qualify as Open Source AI. Each candidate includes an exact variant, SPDX license, evidence record, verifier, and review date. Production publication still requires an OSI-aligned review of the real license, weights, source code, training or processing information, and authorized evidence.

Workload view

Top models by task

Combined-only usage share for each declared or privacy-safe classified task.

General

Classification

11.8% of eligible ranked usage

Coverage: Combined · privacy threshold passed
Classification: Signed metadata first · opt-in classifier second

Combined is an adoption index, not market share.Select OpenRouter or Arc.AI to inspect a source-specific dataset share. Combined percentages are intentionally unavailable.

Versioned evidence

Benchmarks

Versioned benchmark dataset 2026.07.27. Each result records the dataset, scorer, date, and sample size.

RankModelComposite benchmarkContextArc.AI availability

Observed runtime

Fastest models

Arc.AI median time to first token for comparable text requests. OpenRouter latency is not represented.

Latency is source-specific.Select Arc.AI to inspect Arc.AI TTFT measurements. Combined and OpenRouter latency are not inferred.

Language

Languages

Combined-only view for declared natural-language usage.

Code

Programming

Combined-only view for declared programming-language usage.

Prompt shape

Context Length

Combined-only distribution of prompt and completion length.

1K–10K39.4%GLM 5.2

10K–100K31.8%Claude Opus 4.8

100K–1M21.1%MiniMax M3

1M+7.7%MiniMax M3

Agent activity

Tool Calls

Combined-only view of tool-enabled ranked usage.

42.8%of eligible text requests used at least one toolGLM 5.2 · Claude Opus 4.8 · DeepSeek V4 Pro

Multimodal

Images

Combined-only view of eligible multimodal requests.

18.6Mimages in the 30-day snapshotSource and modality coverage shown separately

Privacy-safe application view

Top Apps

Combined-only category aggregates. Named applications require owner opt-in, review, and publication thresholds.

01Coding assistants

33.4%Category aggregate

02Customer support automation

21.8%Category aggregate

03Research & knowledge

16.2%Category aggregate

04Risk and compliance

11.7%Category aggregate

05Other eligible production apps

16.9%Privacy threshold applied

Evidence and limitations

Methodology

Every ranking exposes its source, as-of time, population, calculation, coverage, and privacy status.

Source-separated

OpenRouter and Arc.AI values remain inspectable. Combined never hides a missing or stale source.

Variant-level identity

Aliases map to exact model variants. Unverified mappings stay separate or enter Other.

Privacy thresholded

Arc.AI publishes only eligible aggregate production traffic; customer-identifiable slices are suppressed.

Popularity is not quality

Usage rankings inform discovery. Migration decisions still require workflow-specific evaluation.

OpenRouter and Arc.AI source snapshots remain separately inspectable and are never added as raw token totals.

Arc.AI aggregate snapshot: 2026-07-26 00:00 UTC.

Snapshot 2026.07.27