Telnyx Inference – MiniMax M3 MXFP8 Chat Completions (Large Context)

Choose this endpoint when you need very large context window processing (up to 200,000 characters per message) with MiniMax M3's MXFP8 quantized model via a pay-per-call micropayment model. Prefer this over other Telnyx models when MiniMax M3's specific capabilities (mixture-of-experts architecture, large context) are required, and over standard OpenAI when you want metered per-call billing without a subscription. The upto/metered route is suited for large-context workloads where cost scales with actual usage.

• Payment failure: x402 payment not accepted or insufficient USDC balance
• Context length exceeded: input messages exceed 200,000 characters per message or 32-message limit
• Invalid model field: any value other than MiniMaxAI/MiniMax-M3-MXFP8 will be rejected
• Temperature out of range: values below 0 or above 2 return validation error
• max_tokens out of range: values below 1 or above 8192 return validation error
• Tool list too long: more than 16 tools in the array will be rejected
• Network timeout on very large context payloads