OpenAI GPT-6 Astra Long (>272K context) via mm.family

Choose this endpoint when your input context exceeds 272K tokens and you need GPT-6 Astra specifically — it is the long-context variant designed for inputs that overflow the standard gpt-6-astra endpoint. Prefer it over gpt-6-astra when processing full codebases, lengthy legal documents, or large research corpora. Use the flex tier for cost efficiency, standard for balanced performance, or fast for latency-sensitive workloads. Prefer this over gpt-5.x sibling models when you need the highest capability frontier model for complex long-context reasoning.

• Insufficient USDC balance — payment rejected before completion
• Input token count does not actually exceed 272K threshold — should use gpt-6-astra instead
• Invalid service_tier value — must be one of flex, default, standard, auto, fast, priority
• max_tokens too large for available budget — quoted charge may be rejected
• Model variant string malformed — must be gpt-6-astra-long or gpt-6-astra-long:<level>
• Empty response (zero-length answer) — billed at $0 but counts as a call
• Network timeout on very large context processing under fast tier