DeepSeek V4.1-Flash passes all 10 coding tasks at 40% of Opus cost, but runs 4x slower
创业邦 · wechat · 2026-10-19
DeepSeek released V4.1-Flash-1017 on Oct 17 at unchanged prices, with MIT-licensed weights. Developer Ibragim Badertdinov ran 10 real-world coding tasks three times each:
- All three models (DeepSeek high reasoning, Opus 5.5 low, GPT-6 Astra low) passed. DeepSeek averaged $0.045/task vs $0.1168 for Opus and $0.26 for GPT-6 Astra, but took 2m38s vs 37s and 55s.
- DeepSeek used far more steps (15.3 vs 3.7) and 547K tokens, yet a 97.9% cache-hit rate kept costs down.
- Official demos include 3D scenes and games; one user built a BMW M5 Three.js showcase for $0.38 in 70 minutes, while others report plain landing pages taking 40+ minutes due to over-thinking.
- Lowering the reasoning tier cut time to 1m31s at $0.0257 but failed 3/30 runs; DeepSeek's own Harness passed 29/30, showing toolchain choice matters.
- In the same suite, GPT-6 Luna ($0.0145) and Haiku 5.5 ($0.0208) also passed everything, cheaper than DeepSeek.
Bottom line: strong on small tasks, but large-project reliability and hallucination risks remain unproven.
Related event: DeepSeek V4.1-Flash-1017: 'Minor Update' Tops Open-Weight Leaderboards(12 posts)→
More from coding & agent
- Using Codex to find cute deer in wildlife cam footage and cut a mini video — sch · 2026-10-20
- dumbtv: a no-ads, no-spying custom TV you set up by pointing your coding agent at the repo — Signalman23 · 2026-10-20
- Developer vibecodes a custom UI to manage his fleet of AI agents — _ScottCondron · 2026-10-20
- Salesforce paper: across 2,500 A2A sessions, no agent ever verifies a peer's claims unprompted — rohanpaul_ai · 2026-10-20
- Code-Graph-RAG maps real code dependencies with Tree-sitter and Memgraph, hits 5.3k GitHub stars — JeremyCMorgan · 2026-10-20
- Open-source agent workspace Busabase adds team roles and work auditing — PrajwalTomar_ · 2026-10-20