AI model weekly: Week ending October 25, 2026 — LFM2.5-2.6B lands

Developers should prioritize contingency planning as critical outages affect Anthropic, MiniMax, and Cursor, while reliability metrics show significant degradation for Lovable and xAI Grok. Concurrently, volatile pricing adjustments across Tencent and DeepSeek models alongside new high-context releases from Moonshot AI necessitate immediate cost and architecture reviews.

Outages & incidents

• Fireworks AI — Service Degradation for GLM 5.3 (impact: minor) · status
• Claude (Anthropic) — Elevated errors for Claude Fable 5.1 (impact: major) · status
• Claude (Anthropic) — Elevated errors for Claude Sonnet 5.5 and other models (impact: major) · status
• Lovable — Chat and project changes are currently failing for some users (impact: major) · status
• Runway — Magnific upscalers unavailable (impact: minor) · status
• Claude (Anthropic) — Elevated errors for Claude Fable 5 and Claude Fable 5.1 (impact: major) · status
• MiniMax — Elevated errors on Large Language Models (LLM) (impact: major) · status
• Claude (Anthropic) — Elevated errors for Claude Opus 5.5 (impact: major) · status
• GitHub Copilot — Grok 4.6 and Grok 4.7 Service Issues (impact: minor) · status
• Cursor — Investigating service outage - Grok models unavailable (impact: major) · status
• Claude (Anthropic) — Elevated errors for Claude Fable 5.1, Sonnet 5.5, and Haiku 5.5 (impact: critical) · status
• OpenAI / ChatGPT — Degraded experience with ChatGPT app connections using CIMD (impact: minor) · status

Reliability watch

• Lovable: 104 failed and 55 degraded checks out of 4901
• MiniMax: 45 failed and 108 degraded checks out of 4901
• xAI Grok: 37 failed and 3 degraded checks out of 4902
• OpenAI / ChatGPT: 15 failed and 474 degraded checks out of 4902
• Together AI: 13 failed and 13 degraded checks out of 4901

Price changes

• Qwen3.6 27B (Qwen (Alibaba)): $0.25 / $2.00 → $0.32 / $3.20 (+60%)
• Hy3 (Tencent): $0.13 / $0.53 → $0.0825 / $0.33 (-37%)
• Hy4 preview (Tencent): $0.83 / $2.50 → $0.75 / $2.25 (-10%)
• DeepSeek V4 Flash 0731 (DeepSeek): $0.0422 / $0.17 → $0.0444 / $0.13 (-21%)
• Hy3 (Tencent): $0.0825 / $0.33 → $0.13 / $0.53 (+60%)
• Hy4 preview (Tencent): $0.75 / $2.25 → $0.83 / $2.50 (+11%)
• DeepSeek V4 Flash 0731 (DeepSeek): $0.0444 / $0.13 → $0.0422 / $0.17 (+27%)
• Hy3 (Tencent): $0.13 / $0.53 → $0.0825 / $0.33 (-37%)
• Hy4 preview (Tencent): $0.83 / $2.50 → $0.75 / $2.25 (-10%)
• DeepSeek V4 Pro 0423 (DeepSeek): $0.38 / $0.76 → $0.21 / $0.42 (-45%)
• MiniMax M2.5 (MiniMax): $0.27 / $0.95 → $0.30 / $1.20 (+26%)
• DeepSeek V4 Pro 0423 (DeepSeek): $0.65 / $1.30 → $0.38 / $0.76 (-42%)
• Hy3 (Tencent): $0.0825 / $0.33 → $0.13 / $0.53 (+60%)
• Hy4 preview (Tencent): $0.75 / $2.25 → $0.83 / $2.50 (+11%)
• DeepSeek V4.1 Flash (DeepSeek): $0.30 / $1.20 → $0.02 / $1.00 (-17%)

New models

• LFM2.5-2.6B — Liquid · $0.35 / $0.50 · 128K ctx
• Kimi K3.1 Pro Highspeed — Moonshot AI · $4.00 / $20.00 · 1049K ctx
• Kimi K3.1 Pro — Moonshot AI · $2.00 / $10.00 · 1049K ctx
• Mercury 3 — Inception · $0.10 / $0.35 · 1000K ctx
• MiniMax M3.1 Flash — MiniMax · $0.15 / $0.60 · 1049K ctx

Deprecations & removals

• Trinity Large Thinking retires on 2026-10-26
• Hermes 3 70B Instruct retires on 2026-11-02
• Hermes 3 405B Instruct retires on 2026-11-02
• GLM 4.5V retires on 2026-12-31
• Llama 3.3 Euryale 70B was removed from the catalog
• R1 was removed from the catalog
• WizardLM-2 8x22B was removed from the catalog
• GPT-3.5 Turbo Instruct was removed from the catalog
• GPT-3.5 Turbo was removed from the catalog

This brief is generated automatically from ModelBank's live monitoring data. Want these changes the moment they happen? Get free alerts.