VulcanBench Coding Intelligence Index v1 (VulcanBench CII v1)
We show this table for reference; we do not rank on it.
A post-cutoff software-engineering benchmark with hidden functional tests and regression guards, reported for vendor coding-agent harnesses.
Benchmark score on VulcanBench CII v1 — October 17, 2026
We compile the VulcanBench CII v1 rows from benchmark-owner or independent runs. Claude Opus 5 leads the table at 96.4%, followed by Claude Sonnet 5 (89.2%) and GPT-5.6 Sol (86.5%). We do not use these results to rank models overall.
Claude Opus 5
Anthropic
Claude Sonnet 5
Anthropic
GPT-5.6 Sol
OpenAI
3 modelsCodingCurrentDisplay onlyUpdated October 17, 2026
Benchmark score table (3 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 5Anthropic | 96.4% | Not reported | Closed |
| 2 | Claude Sonnet 5Anthropic | 89.2% | Not reported | Closed |
| 3 | GPT-5.6 SolOpenAI | 86.5% | Not reported | Closed |
Among the reported VulcanBench CII v1 rows, Claude Opus 5 is first at 96.4%. The third row is 9.9 points behind. The broader top-10 range is 9.9 points, so many of the published results sit in a relatively narrow band.
3 models have been evaluated on VulcanBench CII v1. The benchmark falls in the Coding category. VulcanBench CII v1 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About VulcanBench CII v1
Year
2026
Tasks
38 validated post-cutoff repository tasks
Format
Pass@1 with vendor coding-agent harnesses
Difficulty
Mid-band frontier software engineering
CII v1 contains 38 validated tasks from pull requests merged between May and August 2026. The published snapshot measures 37 tasks for Claude Opus 5 with Claude Code CLI and GPT-5.6 Sol with Codex CLI. These are model-plus-vendor-harness results, so we keep them display-only and preserve uncertainty, functional score, task count, and run count in the source snapshot.
Freshness and provenance
Version
VulcanBench CII v1 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
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 VulcanBench CII v1 measure?
A post-cutoff software-engineering benchmark with hidden functional tests and regression guards, reported for vendor coding-agent harnesses.
Which model scores highest on VulcanBench CII v1?
Claude Opus 5 by Anthropic currently leads with a score of 96.4% on VulcanBench CII v1.
How many models are evaluated on VulcanBench CII v1?
3 AI models have published results on VulcanBench CII v1 in the BenchLM catalog.
Compare top models on VulcanBench CII v1
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 5,600+ readers.
One email each week. Unsubscribe anytime.