FrontierSWE 2
Proximal's challenging software engineering agent evaluation (v2).
- Category
- Agentic coding
- Unit
- Percent (%)
- Higher is better
- Comparison groups
- 3total
- Currently available model rankings · Ranked results: 5
- All models (records): Groups: 4 · Entries: 9
Rankings
Values with different source types (official, third-party, paper) or measurement conditions cannot be compared, so rankings are separated by comparison group (the same source, or sources verified to use the same conditions).
Rankings include only models currently available in Claude Code / Codex, excluding legacy models.
official results (Introducing Claude Fable 5.2 and Claude Mythos 5.2)
| Rank | Model | effort | Score | Source cost (metric) | measured/reported by | source |
|---|---|---|---|---|---|---|
| 1 | Claude Fable 5.2 | not stated | 69.9% | — | Anthropic | Introducing Claude Fable 5.2 and Claude Mythos 5.2Retrieved 2026-11-12 · Vendor-reported value. table note: Scores for Fable 5.2, Fable 5.1, and Opus 5.5 show results with safeguards on unless otherwise noted. All scores show each model’s highest-available effort unless otherwise noted.Source excerpt69.9%Location Table, FrontierSWE v2 row: 69.9 |
| 2 | GPT-6 Astra | not stated | 65.5% | — | not stated | Introducing Claude Fable 5.2 and Claude Mythos 5.2Retrieved 2026-11-12 · Vendor-reported value. table note: Scores for Fable 5.2, Fable 5.1, and Opus 5.5 show results with safeguards on unless otherwise noted. All scores show each model’s highest-available effort unless otherwise noted. Values for other vendor models (as listed by Anthropic).Source excerpt65.5%Location Table, FrontierSWE v2 row: 65.5 |
| 3 | Claude Opus 5.5 | not stated | 62.3% | — | Anthropic | Introducing Claude Fable 5.2 and Claude Mythos 5.2Retrieved 2026-11-12 · Vendor-reported value. table note: Scores for Fable 5.2, Fable 5.1, and Opus 5.5 show results with safeguards on unless otherwise noted. All scores show each model’s highest-available effort unless otherwise noted.Source excerpt62.3%Location Table, FrontierSWE v2 row: 62.3 |
official results (Claude Sonnet 5.5 System Card)
| Rank | Model | effort | Score | Source cost (metric) | measured/reported by | source |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5 | Max | 61.9% | — | Proximal | Claude Sonnet 5.5 System CardRetrieved 2026-11-14 · Vendor-reported value. Run by Proximal in its own agent harness. max, 34 tasks, averaged over 5 trials per task. Cost not stated.Source excerpt61.9%Location §8.7 “FrontierSWE v2”, body text, own model score |
official results (Claude Haiku 5.5 System Card)
| Rank | Model | effort | Score | Source cost (metric) | measured/reported by | source |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5 | Max | 43.8% | — | Proximal | Claude Haiku 5.5 System CardRetrieved 2026-11-14 · Vendor-reported value. Run by Proximal in its own agent harness. max, 34 tasks, averaged over 5 trials per task. Cost not stated.Source excerpt43.8%Location §8.6 “FrontierSWE v2”, body text, own model score |
All model rankings (records)
These rankings are recalculated by comparison group including legacy models and models unavailable in Claude Code / Codex. They differ from the rankings above because the set of models differs.
Show rankings including legacy and unavailable models (entries: 9)
official results (Introducing Claude Fable 5.2 and Claude Mythos 5.2)
| Rank | Model | effort | Score | Source cost (metric) | measured/reported by | source |
|---|---|---|---|---|---|---|
| 1 | Claude Fable 5.2 | not stated | 69.9% | — | Anthropic | Introducing Claude Fable 5.2 and Claude Mythos 5.2Retrieved 2026-11-12 · Vendor-reported value. table note: Scores for Fable 5.2, Fable 5.1, and Opus 5.5 show results with safeguards on unless otherwise noted. All scores show each model’s highest-available effort unless otherwise noted.Source excerpt69.9%Location Table, FrontierSWE v2 row: 69.9 |
| 2 | GPT-6 Astra | not stated | 65.5% | — | not stated | Introducing Claude Fable 5.2 and Claude Mythos 5.2Retrieved 2026-11-12 · Vendor-reported value. table note: Scores for Fable 5.2, Fable 5.1, and Opus 5.5 show results with safeguards on unless otherwise noted. All scores show each model’s highest-available effort unless otherwise noted. Values for other vendor models (as listed by Anthropic).Source excerpt65.5%Location Table, FrontierSWE v2 row: 65.5 |
| 3 | Claude Opus 5.5 | not stated | 62.3% | — | Anthropic | Introducing Claude Fable 5.2 and Claude Mythos 5.2Retrieved 2026-11-12 · Vendor-reported value. table note: Scores for Fable 5.2, Fable 5.1, and Opus 5.5 show results with safeguards on unless otherwise noted. All scores show each model’s highest-available effort unless otherwise noted.Source excerpt62.3%Location Table, FrontierSWE v2 row: 62.3 |
| 4 | Gemini 4 Carbon | not stated | 61.2% | — | not stated | Introducing Claude Fable 5.2 and Claude Mythos 5.2Retrieved 2026-11-12 · Vendor-reported value. table note: Scores for Fable 5.2, Fable 5.1, and Opus 5.5 show results with safeguards on unless otherwise noted. All scores show each model’s highest-available effort unless otherwise noted. Values for other vendor models (as listed by Anthropic).Source excerpt61.2%Location Table, FrontierSWE v2 row: 61.2 |
| 5 | Claude Fable 5.1 | not stated | 56.3% | — | Anthropic | Introducing Claude Fable 5.2 and Claude Mythos 5.2Retrieved 2026-11-12 · Vendor-reported value. table note: Scores for Fable 5.2, Fable 5.1, and Opus 5.5 show results with safeguards on unless otherwise noted. All scores show each model’s highest-available effort unless otherwise noted.Source excerpt56.3%Location Table, FrontierSWE v2 row: 56.3 |
official results (Gemini 4 Carbon (Google DeepMind models page, benchmark table))
| Rank | Model | effort | Score | Source cost (metric) | measured/reported by | source |
|---|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon | High | 61.2% | — | Proximal | Gemini 4 Carbon (Google DeepMind models page, benchmark table)Retrieved 2026-11-12 · Vendor-reported value. Google published table. methodology PDF: Gemini 4 Carbon/Argon used the highest Gemini API thinking setting, pass@1, and a single trial. Row-specific source: values from the Proximal official leaderboardSource excerptFrontierSWE 2: Gemini 4 Carbon 61.2 |
| 2 | Gemini 4 Argon | High | 55.0% | — | Proximal | Gemini 4 Carbon (Google DeepMind models page, benchmark table)Retrieved 2026-11-12 · Vendor-reported value. Google published table. methodology PDF: Gemini 4 Carbon/Argon used the highest Gemini API thinking setting, pass@1, and a single trial. Row-specific source: values from the Proximal official leaderboardSource excerptFrontierSWE 2: Gemini 4 Argon 55.0 |
official results (Claude Sonnet 5.5 System Card)
| Rank | Model | effort | Score | Source cost (metric) | measured/reported by | source |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5 | Max | 61.9% | — | Proximal | Claude Sonnet 5.5 System CardRetrieved 2026-11-14 · Vendor-reported value. Run by Proximal in its own agent harness. max, 34 tasks, averaged over 5 trials per task. Cost not stated.Source excerpt61.9%Location §8.7 “FrontierSWE v2”, body text, own model score |
official results (Claude Haiku 5.5 System Card)
| Rank | Model | effort | Score | Source cost (metric) | measured/reported by | source |
|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5 | Max | 43.8% | — | Proximal | Claude Haiku 5.5 System CardRetrieved 2026-11-14 · Vendor-reported value. Run by Proximal in its own agent harness. max, 34 tasks, averaged over 5 trials per task. Cost not stated.Source excerpt43.8%Location §8.6 “FrontierSWE v2”, body text, own model score |