GPT-5.6 Sol input tokens priced at $4 per million after 20% cut

OpenAI published updated pricing for GPT-5.6 Sol on its developer documentation. The model supports reasoning effort levels up to max, 128,000 output tokens, and a 272,000-token cache threshold that triggers 2x input and 1.5x output multipliers. Cache writes incur a 1.25x surcharge. Rate limits scale from Tier 1 at 500 RPM and 500,000 TPM to Tier 5 at 15,000 RPM and 40 million TPM.

Prior GPT-5 family unsuffixed tiers carried higher base rates; the new schedule aligns Sol closer to volume tiers previously reserved for cached or batch workloads. Promotional pricing holds through November 2026. Endpoints now include realtime translation and transcription sessions alongside standard chat completions and fine-tuning.

Operational impact centers on high-context professional workloads. The price delta lowers the marginal cost of 200k-plus token prompts by roughly one-fifth on input and one-third on output, shifting breakeven points for code interpreter and structured output use cases. Tier 3 and above users gain the largest absolute savings under sustained load.

Next observable signals will appear in OpenAI's usage telemetry and any subsequent snapshot releases that lock behavior before the promotional window closes.