GroqCloud supported-models page, re-read 2026-10-17 (production table: gpt-oss public, Llama sales-gated)

Fetched 2026-10-17. Production table, developer plan: llama-3.1-8b-instant and llama-3.3-70b-versatile are both badged Enterprise with Contact Sales in the price and rate-limit columns. llama-3.3-70b-versatile lists vendor speed 280 t/s, context 131072, max completion 32768. openai/gpt-oss-120b is $0.15 input / $0.60 output, vendor speed 500 t/s, context 131072, max completion 65536, 250K TPM / 1K RPM. openai/gpt-oss-20b is $0.075/$0.30, vendor speed 1000 t/s, same context and limits. whisper-large-v3 is $0.111 per hour; whisper-large-v3-turbo is $0.04 per hour. Preview adds qwen/qwen3.8-27b at $0.80/$4.00 and vendor 450 t/s, and openai/gpt-oss-safeguard-20b at $0.075/$0.30. Vendor tok/s figures are list attributes, not a third-party measurement. groq.com/pricing itself no longer renders a rate card.