OpenAI Cuts GPT‑5.6 Sol API Prices by Up to 33%
OpenAI announced on August 21, 2026 that it was cutting GPT‑5.6 Sol API pricing by up to 33% for a three-month promotional window running through November 21. WinBuzzer and multiple API pricing trackers confirmed the new rates within 48 hours of the announcement. For developers and product teams actively calling the Sol API at scale, the cut is material: output token costs dropped from $30 to $20 per million — a one-third reduction — while input tokens fell 20% from $5 to $4 per million.
The announcement lands a week after Anthropic reported its first profitable quarter, and it sharpens an already intensifying price war between the two largest frontier model providers. Here is what changed, what did not, and how to think about the discount if you are building with AI.
The new rates in full
OpenAI updated its pricing page on August 21. The cuts apply to three token types for GPT‑5.6 Sol on the pay-as-you-go API.
| Token type | Old price (per M tokens) | New price (per M tokens) | Change |
|---|---|---|---|
| Input | $5.00 | $4.00 | −20% |
| Output | $30.00 | $20.00 | −33% |
| Cached input | $0.50 | $0.40 | −20% |
The reductions also apply to Codex credits and to eligible ChatGPT Work plans. Subscription tiers — Plus, Pro, and Business — are not part of the promotional window and remain at their previous prices.
What the discount covers — and what it does not
The pricing change is narrower than it might first appear. Three things matter for planning.
It is API and credit spending only. If your AI budget is dominated by ChatGPT subscription seats for a team, this announcement does not change your monthly bill at all. The cut is aimed squarely at developers calling the model programmatically — via the standard API endpoint, via the Responses or Chat Completions APIs, or via Codex for code generation tasks.
It is temporary. OpenAI listed the rate as promotional and valid through at least November 21, 2026. That is roughly three months. The company has not signaled whether the lower price will become permanent or revert. Budget-cycle planning that assumes the new rate holds beyond November carries real risk.
It is Sol-specific. GPT‑5.6 Sol is OpenAI’s highest-capability model in the GPT‑5.6 family. GPT‑5.6 Luna — the lighter, faster, free-tier default — operates at different price points and was not the subject of this announcement. Teams evaluating which model to call should confirm which tier their current integration targets before projecting savings.
How this changes the unit economics for AI-powered products
The output token reduction is the more significant number for most applications. Output tokens are consistently more expensive than input across frontier models because generating text is computationally heavier than reading it. At $30 per million output tokens, Sol was one of the pricier choices for output-heavy tasks like long-form drafting, code generation, or detailed analysis responses. At $20, the calculus shifts.
A product making 10 million output token calls per month — a moderate workload for a small production application — was paying $300 per month on output alone at old prices. At the new rate, that falls to $200. Over three months, that is $300 in savings per 10 million monthly output tokens. For teams running at hundreds of millions of tokens per month, the savings are proportionally larger and worth capturing explicitly.
The input reduction matters more for retrieval-augmented generation (RAG) architectures that pass large context windows with every request. A system routinely sending 50,000 input tokens per call to retrieve context and ground responses sees a 20% per-call cost drop directly from the input price change.
Teams building AI-powered features into their applications — whether through AI coding tools or through direct API integration — should model the impact based on their actual token ratios rather than applying the headline numbers uniformly. The 33% output reduction drives savings for generation-heavy workflows; the 20% input reduction drives savings for context-heavy ones.
Competitive context: where Sol sits after the cut
The price reduction makes GPT‑5.6 Sol meaningfully cheaper than Anthropic’s closest comparable model at standard rates.
| Model | Input (per M tokens) | Output (per M tokens) |
|---|---|---|
| GPT‑5.6 Sol (promotional) | $4.00 | $20.00 |
| Claude Opus 5 (standard) | $5.00 | $25.00 |
At the new rates, Sol is approximately 20% cheaper on both input and output than Claude Opus 5 at standard pay-as-you-go pricing. Whether that differential is decisive depends on factors beyond price: context window size, output quality on your specific task type, latency requirements, and the engineering cost of switching providers if you are already integrated elsewhere.
It is worth noting that Anthropic is not standing still. After reaching profitability ahead of schedule in Q2 2026, Anthropic has more flexibility to respond to competitive price pressure. If Sol’s promotional rate drives meaningful share shifts, a Claude price response before November 21 is plausible. Teams building new integrations for custom software development projects should evaluate both providers’ current rates before locking in a model dependency, as the competitive landscape at this price tier is actively moving.
Should you switch model providers for the discount?
For teams already using GPT‑5.6 Sol in production, the answer is simple: the discount applies automatically. Check your API billing dashboard to confirm the new rate is reflected and update your cost forecasts.
For teams considering a switch from Claude or another provider to capture the Sol discount, the calculation is harder. Switching frontier model providers in a live application is not a configuration change — it requires prompt re-engineering (system prompts, few-shot examples, and output format instructions rarely transfer verbatim between models), regression testing across your use cases, and often significant integration work. A 20–33% cost reduction over three months needs to exceed those one-time migration costs to produce a net benefit.
There is also a compounding risk: if Sol’s pricing reverts to $30 per million output tokens on November 22 and you switched for the promotional rate, you may end up back where you started but with migration costs already spent. The safer position for teams not currently on Sol is to use this window to benchmark Sol seriously against your workloads — and plan a considered migration if the quality holds — rather than rushing a switch purely on price.
What to do before November 21
Three actions make sense regardless of where you currently sit.
- Audit your current token spend by model. Pull three months of API usage from your billing dashboard and separate input from output token costs by model. This gives you a baseline to compare against the promotional Sol rates and project actual savings if you are already on Sol or could reasonably migrate.
- Benchmark Sol on your key tasks now. If you have been deferring a Sol evaluation because of cost, the promotional window removes that barrier. Run your representative prompts against Sol and measure output quality, latency, and format consistency before making any migration decision.
- Watch the OpenAI pricing page in mid-November. OpenAI listed the rate as valid “through at least November 21.” The “at least” language leaves room for an extension. Monitor the pricing page in the two weeks before the deadline to see whether the promotional rate is extended, made permanent, or allowed to revert.
FAQ: OpenAI GPT‑5.6 Sol API price cut
How much did OpenAI cut GPT‑5.6 Sol API prices?
OpenAI reduced GPT‑5.6 Sol API pricing on August 21, 2026. Input tokens dropped from $5 to $4 per million tokens (−20%), output tokens dropped from $30 to $20 per million tokens (−33%), and cached input dropped from $0.50 to $0.40 per million tokens (−20%). The cuts apply to the pay-as-you-go API, Codex credits, and eligible ChatGPT Work plans.
Does the GPT‑5.6 Sol price cut apply to ChatGPT Plus and Pro?
No. The promotional price reduction applies only to the pay-as-you-go API, Codex credits, and eligible ChatGPT Work plans. ChatGPT Plus, Pro, and Business subscriptions remain at their previous prices and are not affected by this discount.
How long does the GPT‑5.6 Sol discount last?
OpenAI has listed the reduced pricing as promotional and valid through at least November 21, 2026. The company has not announced whether the lower rates will become permanent after that date, so teams should plan their AI budgets around the three-month window and monitor OpenAI’s pricing page for any extension or reversion.
How does GPT‑5.6 Sol’s new pricing compare to Claude Opus 5?
At the new promotional rates, GPT‑5.6 Sol costs $4 per million input tokens and $20 per million output tokens. Anthropic’s Claude Opus 5 is priced at $5 per million input tokens and $25 per million output tokens. GPT‑5.6 Sol is now approximately 20% cheaper on both input and output than Claude Opus 5 at standard pay-as-you-go rates.
Should I switch my application from Claude to GPT‑5.6 Sol for the discount?
Not necessarily. Switching AI model providers in a production application has real migration costs: prompt re-engineering, output format changes, regression testing, and integration work. A 20–33% API cost reduction over three months needs to exceed those migration costs to justify a switch. For teams already building on GPT‑5.6 Sol, the discount is straightforward savings. For teams considering their first AI integration, the lower Sol pricing improves the case for choosing OpenAI from the start.