isitsota / categories / Coding
Coding · Oct 13, 2026
Who is SOTA in Coding?
GPT-5 & O4 Mini & DeepSeek R1 0528 & DeepSeek V3.2 Exp & GLM 5.3 Flash & Qwen3 32B
Software engineering skill. Terminal-Bench 4.0 and SWE-bench Pro V2 HARD are the primaries. SWE-bench Verified is shown and ignored for the crown. As of Oct 13, 2026, GPT-5 & O4 Mini & DeepSeek R1 0528 & DeepSeek V3.2 Exp & GLM 5.3 Flash & Qwen3 32B holds Coding. On the cost-quality frontier for…
Standings · pareto
The board
| # | Model | State | Confidence |
|---|---|---|---|
| 1 | AClaude Opus 5.5 Anthropic | Not SOTA · #1 | medium |
| 1 | OGPT-5 OpenAI | Pareto | medium |
| 3 | OO3 Pro OpenAI | Near-SOTA | medium |
| 5 | AClaude Sonnet 5.5 Anthropic | Not SOTA · #5 | medium |
| 6 | OO3 OpenAI | Not SOTA · #6 | medium |
| 11 | OGPT-6.1 Sol OpenAI | Not SOTA · #11 | medium |
| 11 | OGPT-6 Astra OpenAI | Not SOTA · #11 | medium |
| 15 | AClaude Fable 5.1 Anthropic | Not SOTA · #15 | medium |
| 16 | OO4 Mini OpenAI | Pareto | medium |
| 18 | DSDeepSeek R1 0528 DeepSeek | Pareto | medium |
| 20 | AClaude Opus 5 Anthropic | Not SOTA · #20 | medium |
| 21 | DSDeepSeek Chat DeepSeek | Not SOTA · #21 | medium |
| 21 | DSDeepSeek V3.2 Exp DeepSeek | Pareto | medium |
| 24 | OGPT-6 Sol OpenAI | Not SOTA · #24 | medium |
| 25 | OO1 OpenAI | Not SOTA · #25 | medium |
Pareto
Quality against blended price
Evidence recipe
What can move this crown
- Terminal-Bench 4.0 — primary, weight 1
- SWE-bench Pro V2 HARD — primary, weight 1 (display only, cannot crown)
- LMArena Text (style-controlled, coding) — primary, weight 0.6 (display only, cannot crown)
- Terminal-Bench 2.0 — secondary, weight 0.35 (display only, cannot crown)
- Artificial Analysis Coding Index — secondary, weight 0.45
- Aider Polyglot — secondary, weight 0.35
- SWE-bench Verified — display, weight 0 (display only, cannot crown)