AI API Price Changes on December 1, 2026: What Changes
The short answer
Four things hit AI API bills at the turn of the month. As of November 12, 2026:
| Date | What changes | Before | After | Action |
|---|---|---|---|---|
| Dec 1, 2026 | Gemini 4 Argon and Gemini 4 Carbon intro price ends | $2 / $10 per 1M tokens | $4 / $20 | Re-budget, or move routine traffic to a cheaper model |
| Nov 30, 2026 (last day) | GPT-5.6 Sol promotional price ends | $4 / $20 | Not stated by OpenAI (pre-promo list was $5 / $30) | Migrate to GPT-6.1 Sol ($2 / $10) |
| Nov 30, 2026 | claude-sonnet-4-5-20250929 retires on Anthropic-operated platforms | $3 / $15 | Requests fail | Switch to claude-sonnet-5-5 ($2 / $10) |
| Dec 1, 2026 | Cloudflare AI Search billing begins | Free | Usage beyond monthly allowance billed | Check ingestion and query volumes |
Next in line: Gemini 3.8 Flash goes from $0.75 / $3.75 to $1.50 / $7.50 on January 1, 2027.
Gemini 4: every line doubles
Google’s pricing page shows the same schedule for both Gemini 4 models:
| Gemini 4 Argon / Carbon (per 1M tokens) | Through Nov 30 | From Dec 1 |
|---|---|---|
| Standard input / output | $2 / $10 | $4 / $20 |
| Cache reads (implicit only) | $0.10 | $0.20 |
| Batch and Flex input / output | $1 / $5 | $2 / $10 |
| Priority input / output | $3.60 / $18 | $7.20 / $36 |
At $4/$20, Gemini 4 matches Claude Opus 5.5 ($4/$20) and is double GPT-6.1 Sol ($2/$10). Carbon reasons longer than Argon, so per-task costs differ even at the same rate. Context: is Gemini 4 available yet.
What to do: keep Gemini 4 where it wins on your evals; route bulk, latency-tolerant work to Batch or Flex (still $2/$10 after the change); move simple calls to a small model.
GPT-5.6 Sol: the promotion ends November 30
OpenAI cut GPT-5.6 Sol to $4/$20 on August 21, 2026 as a promotion. The pricing page footnote, re-read November 12, says the promotional pricing “is available until November 30, 2026” but gives no December rate. The straightforward move is GPT-6.1 Sol at $2/$10 ($0.10 cached), half the promotional price, which OpenAI positions as its price-performance model.
Claude Sonnet 4.5 retires November 30
Anthropic’s deprecations page lists claude-sonnet-4-5-20250929 as deprecated since September 30, retiring November 30, 2026 on the Claude API, Claude Platform on AWS, Anthropic-managed Google Cloud and Microsoft Foundry. Bedrock and Google-managed Vertex set their own dates. The named replacement, claude-sonnet-5-5, is cheaper at $2/$10, but it has breaking changes (thinking cannot be disabled, forced tool choice returns 400). Migration notes: Claude Sonnet 4.5 retirement: what to use instead. Also on that page: claude-opus-4-5-20251101 is active but retires “not sooner than November 24, 2026”, so watch for a date.
Cloudflare AI Search starts billing December 1
Each account keeps 5 million ingestion tokens, 10 GB-month of storage, 1,000 semantic and 1,000 full-text queries a month free. Beyond that: $0.75 per million ingestion tokens (+$0.50 for image/OCR processing), $2 per GB-month, $0.75 per 1,000 semantic queries and $0.10 per 1,000 full-text queries. A bot answering 100,000 semantic queries a month pays about $74 for queries. More in how to choose an AI knowledge layer.
A two-week checklist
- Search your code and config for
gemini-4-argon,gemini-4-carbon,gpt-5.6-solandclaude-sonnet-4-5. - Recompute monthly cost at the December rates; our current API prices table has both columns.
- Run your eval set on the replacement model before November 30, not after.
- Set billing alerts for December 1 so a doubled rate does not surprise you mid-month.
The general method is in how to budget for AI API price changes.
Last verified: November 12, 2026.