🚨 Breaking
Anthropic — claude-opus-4-8 now GA (2026-05-28)
New top generally-available Anthropic model. Key specs from the release notes: 1M token context window by default, 128k max output tokens, same tool/platform feature set as claude-opus-4-7. Any hardcoded model string in your provider config should be evaluated for upgrade.
→ Anthropic API release notes — May 28, 2026
🗑️ Deprecations
Both retirement deadlines are already past. If you are still referencing either model ID, requests are failing in production right now.
| Model | Retired | Recommended replacement |
|---|---|---|
claude-3-opus-20240229 |
2026-01-05 | claude-opus-4-8 |
claude-2.0 |
2025-07-21 | claude-opus-4-8 |
→ Anthropic model deprecations
💰 Pricing
No pricing changes in today's brief.
🆕 Nouveautés
StepFun Step 3.7 Flash (2026-05-28) — Multimodal MoE: 196B-parameter language backbone paired with a vision encoder for image and video understanding, activating roughly 11B parameters per inference. Available on OpenRouter.
→ openrouter.ai/models/stepfun/step-3.7-flash
goose v1.36.0 (2026-05-27) — TUI mode, diff viewer, goose review local code review command, /goal agent self-evaluation before task completion.
→ GitHub release
Weaviate v1.37.6 (2026-05-27) — No breaking changes. SSB memory limit threshold raised from 80% to 90%; replication abort and backup fixes.
→ GitHub release
llama.cpp (b9370–b9394, 2026-05-27–29) — Hexagon Q4_1 matmul/matvec support, WebGPU legacy constant removal, Gemma 4 audio RMS norm epsilon fix, multimodal n_head_kv default correction.
→ b9394 · b9393 · b9371 · b9370
💡 Conseil du jour
Run grep -rE 'claude-3-opus-20240229|claude-2\.0' . across every repo, config file, and CI secret. Both models were retired months ago — any hit means live production traffic is returning errors today. Swap to claude-opus-4-8 and ship.