Claude Fable 5.1 and a model refresh - Changelo

Claude Fable 5.1 (claude-fable-5-1, released September 1) is the new recommended Claude model on the Claude Code harness — same tier and same base price as Fable 5 ($10 / $50 per million tokens), with cache hits 4x cheaper ($0.25 vs $1).
1M token context, up to 128k output tokens, full reasoning-effort ladder (low → max), defaulting to high.
Fable 5 stays selectable as a legacy model at unchanged prices; Anthropic will not retire it before June 9, 2027.
Sonnet 5 keeps its $2 / $10 launch price — Anthropic made it the standard price instead of raising it to $3 / $15 on September 1.

Grok 4.6 — xAI's recommended model and Grok Build's default (500K context, $2 / $6 per million tokens). Grok 4.5 stays available.
Gemini 3.7 Flash — now the recommended default on the Google Antigravity harness (1M context, $0.75 / $3.75 per million tokens on a 50% promo through December 31).
GLM-5.3 — replaces GLM-5.2 on the Pi harness via OpenRouter or Together, same price, 1M context.
Qwen 3.8 Max — on the Qwen Code harness for orgs with an Alibaba Cloud Model Studio API key (1M context, $2 / $6 per million tokens; not part of the Coding Plan).

DeepSeek V4 Pro (0813) and V4 Flash (0731) — the GA builds, with DeepSeek's peak/off-peak billing: V4 Pro $1.32 / $3.96 at peak and $0.66 / $1.98 off-peak, V4 Flash $0.44 / $1.32 at peak and $0.22 / $0.66 off-peak.
GPT-5.6 — OpenAI's July and August price cuts (Sol $4 / $20, Terra $2 / $12, Luna $0.20 / $1.20 short context) are now reflected in usage estimates. GPT-5.4 was retired from Codex on August 31.
Kimi K2.7 Code — the OpenCode harness now selects it under its correct id (kimi-for-coding/kimi-for-coding; public API id kimi-k2.7-code, $0.95 / $4.00 per million tokens, 262,144 context).