deepseek-v4.1IMAGE
DeepSeek V4.1 Flash across 4 agent setups: 9x speed gap on one task
On the same write-test-fix task, OpenCode with Ollama Cloud took ~6s, dsh with Ollama Cloud ~8s, OpenCode with OpenCode Go ~14s, and Claude Code with Ollama 27-48s, concluding the harness is the bottleneck.
Mohamed Elahmady
@ElA7MaDy_Archived 5 days agoOriginally posted 2026.10.25
Mohamed Elahmady @ElA7MaDy_Originally posted 2026.10.25
Same model. Same task. 4 agent setups. 9x speed gap. DeepSeek V4.1 Flash, write + test + fix: OpenCode + Ollama Cloud: ~6s ⚡ dsh + Ollama Cloud: ~8s OpenCode + OpenCode Go: ~14s Claude Code + Ollama: 27–48s 🐢 Your harness is the bottleneck, not your model. #DeepSeek #Ollama #OpenCode #AIcoding #LLM
x.com/ela7mady_/status/2114192670184603997Archived 2026.10.25 12:00