Scale Labs SWE Atlas – Test Writing (SWE Atlas – Test Writing)

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

A Scale Labs public leaderboard mirrored as display-only reference data. It does not affect BenchLM rankings.

Scale score on SWE Atlas – Test Writing — November 12, 2026 snapshot

We mirror the published scale score view for SWE Atlas – Test Writing. Fable-5.1 (Claude Code) xHigh* leads the public snapshot at 67.0%, followed by Opus 5 (Claude Code) xHigh (62.2%) and Fable-5 (Claude Code) xHigh* (55.6%). We do not use these results to rank models overall.

Fable-5.1 (Claude Code) xHigh*

Anthropic

67.0%

Opus 5 (Claude Code) xHigh

Anthropic

62.2%

Fable-5 (Claude Code) xHigh*

Anthropic

55.6%

24 modelsCodingCurrentDisplay onlyUpdated November 12, 2026 snapshot

Scale score table (24 models)

Scale score results for SWE Atlas – Test Writing
Rank	Model / configuration	Score	Parameters (B)	Open / closed
1	Fable-5.1 (Claude Code) xHigh*Anthropic	67.0%	Not reported	Unknown
2	Opus 5 (Claude Code) xHighAnthropic	62.2%	Not reported	Unknown
3	Fable-5 (Claude Code) xHigh*Anthropic	55.6%	Not reported	Unknown
4	Gemini 3.8 Flash (Mini-SWE-Agent)Google	53.7%	Not reported	Unknown
5	GPT 6 Astra (Codex) xHighOpenAI	50.7%	Not reported	Unknown
6	Opus 4.8 (Claude Code) xhighAnthropic	49.6%	Not reported	Unknown
7	GPT-5.6-Sol (Codex) xHigh*\nOpenAI	45.9%	Not reported	Unknown
8	GPT-5.4 (Codex) xHighOpenAI	44.4%	Not reported	Unknown
9	GPT-5.5 (Codex) xHighOpenAI	42.6%	Not reported	Unknown
10	GLM 5.2 (Mini-SWE-Agent)Z.AI	41.5%	Not reported	Unknown
11	Muse Spark 1.1 (Mini-SWE-Agent) xHighMeta	41.5%	Not reported	Unknown
12	GPT-5.4 (Mini-SWE) xHighOpenAI	40%	Not reported	Unknown
13	GPT 5.3 (Codex) xHighOpenAI	39.0%	Not reported	Unknown
14	Opus 4.7 (Claude Code)Anthropic	38.5%	Not reported	Unknown
15	Opus-4.6 (Claude Code)Anthropic	36.7%	Not reported	Unknown
16	Opus-4.6 (Mini-SWE)Anthropic	36.1%	Not reported	Unknown
17	Sonnet-4.6 (Claude Code)Anthropic	31.8%	Not reported	Unknown
18	Muse SparkMeta	31.1%	Not reported	Closed
19	Gemini-3-Flash (Mini-SWE)Google	30.3%	Not reported	Unknown
20	Gemini-3.1-Pro (Mini-SWE)Google	29.8%	Not reported	Unknown
21	Glm-5 (Mini-SWE)Z.AI	28.7%	Not reported	Unknown
22	Deepseek V4 Pro (Mini-SWE-Agent)Deepseek	27.1%	Not reported	Unknown
23	Kimi-K2.5 (Mini-SWE)Moonshot AI	25.8%	Not reported	Unknown
24	Minimax-M2.5 (Mini-SWE)MiniMax	18.6%	Not reported	Unknown

How to read this leaderboard

Compare the published configurations as complete evaluation systems. The source can combine a base model, agent scaffold, tools, budget, and inference setting in each result.

Operator receipt: 24 sourced rows are currently displayable on this page; the leading published row is Fable-5.1 (Claude Code) xHigh* at 67.0%.

Honest limit: This Scale table is display-only context, not benchmark provenance or a weighted model-only comparison.

SWE Atlas – Test Writing on Scale Labs

How BenchLM shows SWE Atlas – Test Writing

BenchLM mirrors 24 published rows from Scale Labs’ public SWE Atlas – Test Writing leaderboard, captured on November 12, 2026 snapshot.

The table is display only. It is useful context for a published agent or model configuration, but it does not enter BenchLM’s overall or category rankings.

Snapshot

24 published rowsScale Labs sourceDisplay only

SWE Atlas – Test Writing on Scale Labs

The published SWE Atlas – Test Writing snapshot places Fable-5.1 (Claude Code) xHigh* first at 67.0%. The third row is 11.4 points behind. The broader top-10 range is 25.6 points, so the table still separates the published systems.

24 models have been evaluated on SWE Atlas – Test Writing. 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. SWE Atlas – Test Writing is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About SWE Atlas – Test Writing

Year

2026

Tasks

24 published rows

Format

Published Scale leaderboard score

Difficulty

External agent and model evaluation

BenchLM mirrors 24 published rows from the SWE Atlas – Test Writing public table captured on November 12, 2026 snapshot.

Scale Labs leaderboard Public benchmark source

Freshness and provenance

Version

SWE Atlas – Test Writing 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay 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 SWE Atlas – Test Writing measure?

A Scale Labs public leaderboard mirrored as display-only reference data. It does not affect BenchLM rankings.

Which model leads the published SWE Atlas – Test Writing snapshot?

Fable-5.1 (Claude Code) xHigh* currently leads the published SWE Atlas – Test Writing snapshot with 67.0% scale score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on SWE Atlas – Test Writing?

The November 12, 2026 snapshot snapshot contains 24 AI models.

Last updated: November 12, 2026 snapshot · 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 7,000+ subscribers.

One email each week. Unsubscribe anytime.