2025 AI Model Status and Updates | Generated by AI
Here’s a breakdown of which models in your mapping are now outdated / superseded within their series as of the end of 2025:
✅ Outdated or Superseded
These are models that have newer versions or have been deprecated/retired by their own developers:
Anthropic (Claude)
- “claude-opus” → anthopic/claude-opus-4.1: Older Opus 3-series is deprecated, and the current line is Opus 4 and 4.1. If you were mapping to Opus 3 previously, that’s outdated. (Claude Docs)
- “claude-sonnet” → anthropic/claude-4.5-sonnet: The older 3-series Sonnet (e.g., Claude 3-Sonnet and Sonnet 3.5) is retired or deprecated, replaced by Sonnet 4.x. (Claude Docs)
Mistral
- “mistral” → mistralai/mistral-medium-3.1: Mistral has newer generations (like the broader Mistral 3 family with Large 3, Medium 3.1, Small 3.2 etc.). Medium-3.1 itself is current, so it replaces older Medium-3.0 variants. (Mistral AI Documentation)
Other Potentially Outdated (less clear):
- “qwen” → qwen/qwen3-coder: The Qwen family is evolving quickly, with newer variants beyond a simple 3-coder base (e.g., Qwen3 Pro, Qwen3 Ultimate etc. appear in releases; see industry news). So a plain “qwen3-coder” might be not the latest in that series. (WIRED)
🟢 Still Current / Not Clearly Outdated
These models represent recent or flagship versions and don’t have widely known newer successors yet:
- “gpt” → openai/gpt-5.2: OpenAI’s GPT-5.2 is currently the latest major public GPT version (GPT-5.1 was released before). (ChatHub)
- “grok-code” → x-ai/grok-code-fast-1: Latest Grok fast coding variant; no widely publicized replacement yet.
- “grok-fast” → x-ai/grok-4.1-fast: Grok 4.1 is current in the 4-series as far as mainstream references show.
- “glm” → z-ai/glm-4.7: Current GLM 4-series model; no widely known successor yet.
- “deepseek” → deepseek/deepseek-v3.2: That v3.2 lineage is the latest published by DeepSeek (with some experimental variants). (DeepSeek API Docs)
- “kimi” / “kimi-thinking” → moonshotai/kimi-k2 / k2-thinking: Kimi K2 is the current generation; no newer MoonshotAI model yet widely referenced.
- “minimax” → minimax/minimax-m2: MiniMax still appears current with the M2 baseline.
🧾 Summary (Quick View)
Outdated / replaced:
claude-opus(if it pointed at older Opus 3)claude-sonnet(if it pointed at older Sonnet 3/3.5)
Current:
gpt(GPT-5.2)grok-code/grok-fastglmdeepseekmistral(Medium 3.1)qwen(present model, though the family is evolving)kimi/kimi-thinkingminimax
If you want, I can also tell you which of these models have even newer unreleased or beta versions announced or specify their approximate training cutoffs and use-case strengths (reasoning, coding, agent workflows, etc.).