EEBench V1 core corpus (EEBench)
We show this table for reference; we do not rank on it.
An electrical-engineering agent benchmark maintained by atopile.
EEBench score on EEBench — November 12, 2026 snapshot
We mirror the published eebench score view for EEBench. Gemini 4 Carbon leads the public snapshot at 83.9%, followed by Claude Fable 5.2 (77.3%) and Claude Opus 5.5 (75.0%). We do not use these results to rank models overall.
Gemini 4 Carbon
Gemini CLI
Claude Fable 5.2
Anthropic
Claude Code
Claude Opus 5.5
Anthropic
Claude Code
32 modelsCodingCurrentDisplay onlyUpdated November 12, 2026 snapshot
EEBench score table (32 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Gemini 4 CarbonGoogleGemini CLI | 83.9% | Not reported | Closed |
| 2 | Claude Fable 5.2AnthropicClaude Code | 77.3% | Not reported | Closed |
| 3 | Claude Opus 5.5AnthropicClaude Code | 75.0% | Not reported | Closed |
| 4 | GPT-6 AstraOpenAICodex CLI | 69.3% | Not reported | Closed |
| 5 | Claude Opus 5.5AnthropicClaude Code | 68.9% | Not reported | Closed |
| 6 | Claude Sonnet 5.5AnthropicClaude Code | 67.2% | Not reported | Closed |
| 7 | Claude Haiku 5.5AnthropicClaude Code | 67.1% | Not reported | Closed |
| 8 | Gemini 4 ArgonGoogleGemini CLI | 67.1% | Not reported | Closed |
| 9 | Grok 4.7xAIGrok CLI | 64.0% | Not reported | Closed |
| 10 | GPT-6.1 SolOpenAICodex CLI | 63.6% | Not reported | Closed |
| 11 | Muse Wonder 1.0MetaMuse Code | 62.7% | Not reported | Unknown |
| 12 | Claude Opus 5AnthropicClaude Code | 61.6% | Not reported | Closed |
| 13 | Grok 4.6xAIGrok CLI | 57.1% | Not reported | Closed |
| 14 | Claude Fable 5.1AnthropicClaude Code | 56.4% | Not reported | Closed |
| 15 | GPT-6 SolOpenAICodex CLI | 56.3% | Not reported | Closed |
| 16 | Gemini 3.8 FlashGoogleGemini CLI | 55.4% | Not reported | Closed |
| 17 | Claude Fable 5AnthropicClaude Code | 54.3% | Not reported | Closed |
| 18 | Claude Opus 4.8AnthropicClaude Code | 51.4% | Not reported | Closed |
| 19 | Grok 4.5xAIGrok CLI | 50.9% | Not reported | Closed |
| 20 | Mistral Large 4MistralMistral Vibe | 50.1% | Not reported | Open |
| 21 | Muse Spark 1.3MetaMuse Code | 49.4% | Not reported | Closed |
| 22 | Gemini 3.1 ProGoogleGemini CLI | 46.3% | Not reported | Closed |
| 23 | Muse Ember 1.0MetaMuse Code | 42.8% | Not reported | Unknown |
| 24 | Claude Sonnet 5AnthropicClaude Code | 40.3% | Not reported | Closed |
| 25 | Gemini 3.6 FlashGoogleGemini CLI | 37.4% | Not reported | Closed |
| 26 | Claude Opus 4.8AnthropicClaude Code | 35.2% | Not reported | Closed |
| 27 | Gemini 3.5 FlashGoogleGemini CLI | 34.3% | Not reported | Closed |
| 28 | GPT-6 LunaOpenAICodex CLI | 31.1% | Not reported | Closed |
| 29 | Claude Sonnet 4.6AnthropicClaude Code | 20.4% | Not reported | Closed |
| 30 | Gemini 3 FlashGoogleGemini CLI | 19.5% | Not reported | Closed |
| 31 | Grok Build 0.1xAIGrok CLI | 18.0% | Not reported | Closed |
| 32 | Claude Haiku 4.5AnthropicClaude Code | 3.5% | Not reported | Closed |
How we show EEBench
We mirror the official EEBench V1 leaderboard generated on November 12, 2026 snapshot. The snapshot preserves all 32 published model-and-scaffold configurations together with uncertainty, attempt, cost, token, step, and duration fields.
EEBench is display only. The public results combine a model with scaffolds such as Claude Code, Grok CLI, Gemini CLI, API tools, or OpenRouter, so we keep them outside weighted model-only rankings.
Snapshot
The published EEBench snapshot places Gemini 4 Carbon first at 83.9%. The third row is 8.9 points behind. The broader top-10 range is 20.3 points, so the table still separates the published systems.
32 models have been evaluated on EEBench. The benchmark falls in the Coding category. EEBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About EEBench
Year
2026
Tasks
Electrical-engineering design tasks
Format
Agent task score
Difficulty
Professional electrical engineering
The public leaderboard exposes score, cost, and output-token fields. BenchLM stores the Grok 4.6 card comparison as display-only system evidence.
Freshness and provenance
Version
EEBench 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 EEBench measure?
An electrical-engineering agent benchmark maintained by atopile.
Which model leads the published EEBench snapshot?
Gemini 4 Carbon currently leads the published EEBench snapshot with 83.9% eebench score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on EEBench?
The November 12, 2026 snapshot snapshot contains 32 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 7,000+ subscribers.
One email each week. Unsubscribe anytime.