docs: Add DeepSeek V4 Pro/Flash support to agent-model-matching guide

## Summary [DeepSeek V4 Preview](https://api-docs.deepseek.com/news/news260424) was released on **April 24, 2026** (2 days ago), shipping two new models: | Model | API ID | Params (Total/Active) | Context | Pricing (Input/Output per 1M) | |-------|--------|-----------------------|---------|-------------------------------| | **DeepSeek-V4-Pro** | `deepseek-v4-pro` | 1.6T / 49B | 1M | $1.74 / $3.48 | | **DeepSeek-V4-Flash** | `deepseek-v4-flash` | 284B / 13B | 1M | $0.14 / $0.28 | Both models are **OpenAI ChatCompletions API compatible** (base URL: `https://api.deepseek.com`), support thinking/non-thinking modes, tool calls, JSON output, and FIM completion. V4-Pro achieves agentic coding SOTA among open-source models, rivaling top closed-source models. --- ## Proposed Changes ### 1. `docs/guide/agent-model-matching.md` — New "DeepSeek Family" section Add a new model family entry under the existing families (Claude, GPT, Gemini, etc.): ```markdown ### DeepSeek Family OpenAI-compatible, strong reasoning at aggressive pricing. Best for cost-sensitive delegation or as an OpenAI fallback. Integrates via the `openai` provider path with `base_url` override or as a dedicated gateway entry. | Model | Strengths | |-------|-----------| | **DeepSeek V4 Pro** (`deepseek-v4-pro`) | Agentic coding SOTA among open models. 1.6T MoE, rivals top closed-source models for deep reasoning. Good for `deep` / `unspecified-high` categories. | | **DeepSeek V4 Flash** (`deepseek-v4-flash`) | Fas

Posted by RouHim on Apr 26, 2026, 10:10 AM

Tags

• pricing (Logistics)

Related issues

• A lot of the rare books seem to be books that are most relevant to specific professions within specific countries.
• It's cute, looks nice, and bravo for shipping something, but FZ is too hyped and too expensive for what it is.
• I'd really like to see a plug-in hybrid pickup where I can do 10-20 miles on a charge without needing the gas engine to even kick on.