Claude Code with Third-Party Models: Configuration and Official Support Boundary
What a third-party Anthropic-compatible gateway workflow can do with Claude Code, and the official Anthropic support boundary that must never be overstated.
Claude Code with Third-Party Models: Configuration and Official Support Boundary
Claude Code with Third-Party Models: Configuration and Official Support Boundary is documented here with numbers verified directly against the HuggingFace API and the raw config.json of the corresponding repository, sampled in November 2026. No benchmark score, price, or repository name is invented; anything the config cannot prove is labelled as needing measurement.
The 30-second version
Anthropic officially supports Claude models; a third-party gateway is not the same thing. Only Chinese open models are served through that compatible endpoint
Verified specification
| Item | Value | Notes |
|---|---|---|
| Official Anthropic support | Claude models | Not non-Claude models |
| Third-party gateway workflow | User-side configuration | Not an official feature |
| Anthropic-compatible endpoint | EXAIV /api/ai/v1/anthropic/messages | OpenAI-compatible chat endpoint also exists |
| Callable IDs | glm-5.3-flash, kimi-k3, deepseek-v4.1-flash | Follow the live catalog |
| Environment variables | Check the official Claude Code docs | Do not invent names |
| Streaming | Standard OpenAI-compatible format | Works on the gateway endpoint |
| Licence and terms | Vendor-specific | Anthropic terms for Claude Code |
| Support boundary | Read the official documentation | Never overstate |
What the numbers actually confirm
The distinction between verified fact and claim matters most for multimodal wording. A vision_config and an image_placeholder in the official config confirm an image-capable conditional-generation architecture; they do not by themselves prove video generation, a specific request payload, or any benchmark result. The HuggingFace tag image-text-to-text describes the input direction. Always treat the official model card as the final source of truth.
Deployment and configuration reality
Memory planning follows the same rule for every trillion-parameter MoE: compute scales with active parameters while memory scales with total parameters. A 1T model at 4-bit still needs roughly 500GB for weights alone, and a 2.78T model is larger still, so these are multi-GPU server workloads rather than single consumer cards. For a single card, look at the lightweight families. The local path is vLLM or SGLang with official weights, and community conversions should be checked for source and quantization label first.
Honest boundaries
On EXAIV the live /models catalog is authoritative: only kimi-k3 is callable today, through the OpenAI-compatible endpoint at https://exaiv.com/api/ai/v1. Integration changes only base_url, api_key and model; streaming and tool calling keep the standard OpenAI format. Prices always follow the live /pricing page and are never hard-coded here. Keep a log of model ID, token counts and latency so a later model swap has the same baseline.
Scenario selection
| Situation | Pick |
|---|---|
| Use Claude Code as documented | Official Anthropic setup |
| Experiment with a third-party gateway | User-side configuration only |
| Need guaranteed compatibility | Official documentation |
| Call a Chinese open model | EXAIV OpenAI-compatible endpoint |
| Need verified third-party support | Ask the gateway provider |
| Check a specific model ID | The live /models catalog |
| Publish a setup guide | State the support boundary explicitly |
| Confirm pricing | The live /pricing page |
FAQ
Which EXAIV model ID should I use?
Are the numbers in this article verified?
Does the config prove video generation?
Do I need a new SDK?
Where do I check prices?
Can I use local weights?
Related reading
- Kimi K3 API guide — first API call
- hy4 model selection guide — sizing and selection
- GLM model pages — catalog and specs
- Claude Code third-party API — Anthropic compatible endpoint
- pricing — live credit packs
Sources
Parameter and file data sampled in November 2026 and subject to change.