DeepSeek V4.1 Flash: benchmark results
DeepSeek's open-weight V4.1 Flash (September 2026), a 552B-parameter multimodal MoE whose causal encoder-decoder design activates 8B parameters at prefill and 16B at decode. Provider: DeepSeek. Released 2026-09-10. Access: Open.
Aggregate Rating 1845 ± 6, rank #35 of 1635 rated models, from 590 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AI Coding Daily (OpenCode) - React-TS Code Quality | 18.67 | React-TS Code Quality (max 20) points, LLM-judged rubric sco | 100 |
| AgenticBBO Frontier - Bigblue1 Chip Placement | 0.67 | Normalized anytime/final score (0-1) | 100 |
| Clerk LLM Leaderboard | 92 | Avg score (%) | 100 |
| Codeforces Rating (DeepSeek V4.1 report) | 3471 | Codeforces rating, DeepSeek V4.1 report, Max effort (self-re | 100 |
| EviRepair Study - Codex - PatchEval Go | 74.7 | Resolved (%) | 100 |
| EviRepair Study - Codex - PatchEval JavaScript | 79.22 | Resolved (%) | 100 |
| EviRepair Study - Codex - PatchEval Overall | 78.26 | Resolved (%) | 100 |
| EviRepair Study - Codex - PatchEval Python | 81.43 | Resolved (%) | 100 |
| EviRepair Study - Codex - SEC-bench CVE | 64.5 | Resolved (%) | 100 |
| EviRepair Study - Codex - SEC-bench OSS-Fuzz | 90 | Resolved (%) | 100 |
| EviRepair Study - Codex - SEC-bench Overall | 73 | Resolved (%) | 100 |
| EviRepair Study - EviRepair - PatchEval Go | 91.57 | Resolved (%) | 100 |
Interactive version: theaggregate.ai/model?slug=deepseek-v4-1-flash · How It Works · Data refreshed daily, snapshot 2026-10-15.