DeepSeek's 1017 update and the model name that moves under you
DeepSeek swapped the model behind deepseek-flash on 17 October, and legacy names followed. Why aliases change agent behaviour, and how to pin and re-test.
Choosing which model answers a call: rules, advisors and fallbacks.
DeepSeek swapped the model behind deepseek-flash on 17 October, and legacy names followed. Why aliases change agent behaviour, and how to pin and re-test.
Three coding-relevant models landed within ten days. Rather than picking one default, map task types to model tiers and check vendor benchmarks with care.
Codex 0.161.0 switched its default model to GPT-6.1 Sol and made Daybreak cyber routing opt-in. What changed, what it costs, and what to pin in CI and config.
5 min read
OpenAI's GPT-6.1 Sol lists at $2/$10 per million tokens against Astra's $10/$50. What that changes for default coding models, caching and harness cost.
Anthropic's Haiku 5.5 makes high-volume subagent work cheap enough to measure. When to route there, and when Sonnet still leads.
Muse Wonder launched with both API prices and subscription-only access. How the two pricing models differ, and what that means for controlling AI spend.
3 min read
Turn on the SpanDock model proxy, give each person a budget, choose what happens at the limit, and see what was spent and saved.