Driveline team's model lineup: Opus 5.5 still king for coding, Carbon 4 for reviews
ivan_bezdomny · x · 2026-11-18
A Driveline Technology employee shared how the team assigns AI models by function: GPT-6.1-Sol for IT ops/infra, Opus 5.5 for coding and design, Carbon 4 for technical/adversarial/research reviews, Fable 5.2 for orchestration, and Daybreak, DS4 Flash Abliterated, Mythos 5.2 for security.
He says they dropped Sol/Astra for code — too many issues, overly defensive with "contracts" and "gates," useless tests. Opus 5.5 remains king for piecemeal and greenfield work, and he doubts incremental OpenAI updates can match it short of a GPT-7. He adds that Carbon stands out in research reviews for being far more independent in thought and reasoning direction, where the Opus/Astra adversarial pipeline falls short.
More from coding & agent
- Sentry founder jokes about MCP as HN thread declares it the standard for AI agent tooling — zeeg · 2026-11-18
- Self-hosted Qwen3.8-27B + pi LoRA matches Opus 5.5 on a real bug fix in ~93s — yb2698 · 2026-11-18
- Astra 6.2 hailed as best computer-use model after flawless MS Paint demo — gabrielchua · 2026-11-18
- Solo dev built a 20k-user ham radio app and a multiplayer game almost entirely with Claude — GreyFoxSolid · 2026-11-18
- Asking Claude to Code a Particle Simulation: 4.2k Balls Sync Into Perfect Spirals — mickeyxfriedman · 2026-11-18
- Qwen 3.8 Flash Next praised for pleasant, human-like coding updates — user hesitates to upgrade away — catplusplusok · 2026-11-18