BridgeBench
We show this table for reference; we do not rank on it.
BridgeBench rates 25 models across nine software-building axes, including reasoning, front end, back end, security, speed, and cost. Its overall rating is the equal-weight mean of those axes.
Overall editorial rating on BridgeBench — September 23, 2026
We mirror the published overall editorial rating view for BridgeBench. Claude Opus 5.5 leads the public snapshot at 760, followed by GPT-6 Astra (734) and Claude Fable 5.1 (711). We do not use these results to rank models overall.
Claude Opus 5.5
Anthropic
GPT-6 Astra
OpenAI
Claude Fable 5.1
Anthropic
25 modelsCodingCurrentDisplay onlyUpdated September 23, 2026
Overall editorial rating table (25 models)
ScoreHow to read this leaderboard
Compare the source's software-building assessments as one editorial view. The overall mean includes speed and cost alongside capability axes, so a higher rating is not a pure capability claim.
Operator receipt: 25 sourced rows are currently displayable on this page; the leading published row is Claude Opus 5.5 at 760.
Honest limit: BridgeBench does not publish its proprietary prompts, task sets, or detailed rubrics. Confidence varies by model and axis, and the source says small rating gaps do not prove a meaningful difference on every task. These ratings do not enter BenchLM overall or category rankings.
How we show BridgeBench
We captured 25 overall ratings from the public BridgeBench leaderboard on September 23, 2026. BridgeBench averages nine axes, including reasoning, front end, back end, security, speed, and cost, with equal weight.
These are BridgeBench editorial assessments expressed as Elo-style points. They are not head-to-head match records or pass rates from a shared test suite. The source keeps its prompts, task sets, and detailed rubrics private.
We show this table for reference only. Its ratings do not enter our overall or category rankings.
Snapshot
The published BridgeBench snapshot places Claude Opus 5.5 first at 760. The third row is 49 score units behind. The broader top-10 range is 177 score units, so the table still separates the published systems.
25 models have been evaluated on BridgeBench. 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. BridgeBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About BridgeBench
Year
2026
Tasks
No public task count
Format
Editorial Elo-style rating
Difficulty
Software-building assessments
The published ratings combine editorial judgment with available performance evidence. BridgeBench converts comparative assessments to Elo-style points; they are not head-to-head match results or pass rates from one shared test suite. We mirror the 25 published overall ratings as display-only context.
Freshness and provenance
Version
September 2026 editorial ratings
Refresh cadence
Manual snapshot
Staleness state
Current
Question availability
Prompts, task sets, and detailed rubrics private
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 BridgeBench measure?
BridgeBench rates 25 models across nine software-building axes, including reasoning, front end, back end, security, speed, and cost. Its overall rating is the equal-weight mean of those axes.
Which model leads the published BridgeBench snapshot?
Claude Opus 5.5 currently leads the published BridgeBench snapshot with 760 overall editorial rating. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on BridgeBench?
The September 23, 2026 snapshot contains 25 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 2,000+ readers.
One email each week. Unsubscribe anytime.