Inference APIs
Reference/Model status

openai/gpt-oss-20b

Together retired it on 2026-09-15 (Serverless; dedicated endpoints unaffected). Announced: undated (Together publishes removal dates only). Their replacement: Qwen/Qwen3.5-9B.

Served here as the same model at $0.04 / $0.19 per 1M tokens. Base URL https://api.inferenceapis.com/v1, model string unchanged. Model page

JSON: curl "https://inferenceapis.com/reference/model-status?id=openai%2Fgpt-oss-20b&format=json"

Try: llama-3.3-70b-versatile · meta-llama/Llama-3.3-70B-Instruct-Turbo · deepseek-ai/DeepSeek-V4-Flash-0731 · openai/gpt-oss-20b · mixtral-8x7b-32768 · moonshotai/Kimi-K2.6 · qwen/qwen3-32b

What it checks

  • Retired? Groq's and Together's deprecation pages, re-read daily; a row gives the shutdown date, what the provider recommends, and the scope (free tier, serverless, all tiers).
  • Served here? Whether the id is one of the 164 ids the gateway accepts: the same model under the same id, or an alias routed to the closest model, with the current price.

Related: Groq alternative, Together alternative, error reference, switching providers.