OpenAI cut the price of its flagship GPT-5.6 Sol model by more than 20% on August 21, 2026, making it cheaper than Anthropic's Claude Opus 5 for the first time since Sol launched. Input pricing dropped from $5 to $4 per million tokens, while output pricing fell from $30 to $20 per million tokens. The catch: this promotional pricing expires on November 21, 2026, and OpenAI has not announced whether rates will return to their original levels.

Why Is This Price War Happening Now?

OpenAI is defending its market share on two fronts simultaneously. On the high end, Anthropic has been gaining ground with enterprises; a July 2026 analysis by expense management platform Ramp found that 43.5% of US enterprises paid Anthropic token fees, compared with 39.7% for OpenAI. On the low end, Chinese AI models like DeepSeek's V4 Pro and Moonshot AI's Kimi K3 are undercutting both American companies with dramatically lower prices, forcing major companies like Airbnb and Siemens to test cheaper alternatives.

The timing also matters: OpenAI's Chief Financial Officer Sarah Friar has indicated the company is working toward a 2027 initial public offering, while Anthropic has already filed confidentially for its own listing. For OpenAI, a price cut signals momentum to potential investors and helps stabilize revenue before going public.

How Does Sol's New Pricing Compare to Claude?

The pricing shift fundamentally changes the flagship comparison. Before August 21, Sol tied Claude Opus 5 on input pricing ($5 per million tokens) but lost on output ($30 versus $25). Now Sol beats Opus 5 on both metrics: $4 input versus $5, and $20 output versus $25.

The gap widens even further when comparing Sol to Anthropic's Claude Fable 5. Sol now costs just 40% of what Fable 5 charges on both input and output, a significantly wider margin than the one it closed against Opus 5.

What Should Enterprises Do About This Temporary Discount?

• Evaluate the November Expiration Date: The promotional pricing is locked in only through November 21, 2026. If the discount lapses as scheduled, Sol's output price will jump 50%, from $20 back to $30 per million tokens. Any production migration completed in October could face a price hike by December.
• Consider Your Existing Investment: If your team has already built workflows, prompts, or tooling around Claude Opus 5, a 20% price gap that could vanish in three months may not justify rebuilding your entire pipeline. The switching cost often outweighs the short-term savings.
• Look Beyond Flagship Models: The real cost savings often come from moving down a tier rather than picking the marginally cheaper flagship. Anthropic's Haiku 4.5 and Sonnet 5 models, or OpenAI's Terra and Luna variants, may offer better value for routine classification, extraction, or drafting work than either flagship option.

Is Anthropic Responding to This Price Cut?

As of the publication date, Anthropic has not changed Claude Opus 5's pricing in response to OpenAI's move. Claude Opus 5 remains at $5 input and $25 output per million tokens. Anthropic's silence suggests confidence in its model's quality and enterprise adoption, even at a higher price point. The Ramp data showing 43.5% of US enterprises using Anthropic in July indicates that cost alone is not driving purchasing decisions; many teams value Claude's performance enough to pay a premium.

However, the competitive pressure is real. OpenAI's aggressive pricing on Sol, combined with the wave of cheaper Chinese models, suggests the AI industry is entering a new phase where cost efficiency and inference speed matter as much as raw capability. Anthropic may eventually respond, but for now, the company appears to be betting that enterprise customers will stick with Claude based on quality and reliability rather than price alone.

What Happens After November 21?

OpenAI has committed to the new pricing through at least November 21, 2026, but has said nothing about what happens after that date. If the promotion lapses on schedule, Sol's output pricing will return to $30 per million tokens, a 50% increase over the promotional rate. This uncertainty creates a real planning challenge for enterprises: building a cost model around $20 output pricing is essentially making a bet on a number with an expiration date stamped on it.

The broader lesson is that promotional pricing in the AI market is becoming a standard competitive tactic. OpenAI used the same framing when it cut prices on its lower-tier models Terra and Luna in July 2026, describing the cuts as passing along efficiency gains to customers. But the pattern suggests these are tactical moves designed to test market resistance and defend share, not permanent repricing.

"As we get more efficient, we are passing the savings on to you," said Scott Rosecrans, Vice President of Strategic Pursuits at OpenAI, discussing the July price cut.

Scott Rosecrans, Vice President of Strategic Pursuits, OpenAI

For enterprises making long-term AI infrastructure decisions, the key metric to track is not this week's price, but what OpenAI announces on November 21. That date will reveal whether the company views this discount as a temporary competitive maneuver or the beginning of a sustained shift in its pricing strategy. Until then, any production migration built around Sol's promotional rates carries real financial risk.