Model

OpenAI cuts flagship pricing 20% to match Anthropic as heavy users call GPT-6.2 Astra unusable

OpenAI cut its flagship's price to match Claude Fable 5.2's rate, and the users who run models hardest rejected it, 24 hours before Anthropic priced a $1.4 trillion IPO on the assumption the quality gap holds. The independent labs say the gap is real, for now.

Vincent JiangVincent Jiangthree-minute read
Share
Sam Altman, in a dark suit, shakes hands with Indian Prime Minister Narendra Modi
Sam Altman, OpenAI's chief executive, with Prime Minister Narendra Modi in New Delhi in February 2026.

OpenAI$1.18T — OpenAI, private, latest valuation $1.18T shipped two models on November 17: GPT-6.2 Astra, its flagship for professional work in science, law and finance, and GPT-6.2 Sol, the cheaper everyday model 12. Astra's API price fell 20% to $8 per million input tokens and $40 per million output tokens, exactly the rate Anthropic$350B — Anthropic, private, latest valuation $350B charges for Claude Fable 5.2 12.

The users who run models hardest were not buying it. Investor Matt Shumer called Astra "essentially unusable"; developer Theo Browne called it "the worst model OpenAI has ever released" after seven days of delegated work left his experimental Rust compiler rewrite running 56% slower 1. Wharton professor Ethan Mollick called it the first OpenAI flagship since GPT-4.5 that is not a clear step forward 1.

The IPO prices a premium OpenAI just undercut

One day later, Anthropic was set to price the largest IPO in years: 825 million shares at $70 to $78, more than $64 billion raised, a valuation near $1.4 trillion, with orders around seven times the shares on offer 36.

Trading under the ticker ANT begins Thursday. That price rests on premium-model pricing power, the kind an $8/$40 flagship charges for 2. OpenAI's counter is cost per task: it claims Astra beats Fable 5.2 on the professional-work benchmark Agents' Last Exam by 5.5 points at 5% of the cost 2. The falling price of intelligence we flagged in September is now aimed directly at the premium Anthropic shareholders are paying for 7.

The independent tests still put Fable ahead

Independent testing says the quality gap is still real. On Artificial Analysis' Intelligence Index, Astra scores 62, second to Fable 5.2's 65 24. Coding flatlined: Astra matched its predecessor in the Coding Agent Index and Sol dropped three points 2. At maximum effort, Astra guesses instead of abstaining on 59% of hard factual questions, per Artificial Analysis' hallucination measure 2. Epoch AI ties Astra with Fable at 174 on its capabilities index, but on real research tasks drawn from its own work, neither model met Epoch's employee standard 5.

Claude Fable 5.2 still tops the independent intelligence index

020406080Claude Fable 5.2GPT-6.2 AstraGemini 4 CarbonClaude Opus 5.5GPT-6.2 SolGPT-6 Astra5762
Data
Intelligence Index score
Claude Fable 5.265
GPT-6.2 Astra62
Gemini 4 Carbon60
Claude Opus 5.559
GPT-6.2 Sol57
GPT-6 Astra57
Scores on Artificial Analysis' Intelligence Index v5.0.1, which runs each model through a common battery of reasoning, coding and knowledge evaluations; higher is better. GPT-6.2 Astra highlighted. Sources: Artificial Analysis (accessed November 18, 2026); The Decoder.2,4

Astra wins where the work is cheap

Astra's supporters have numbers too. It leads Terminal-Bench-Science, the AA-3D design ranking and robotics, and Artificial Analysis says OpenAI owns the cost-performance frontier, the line where no model is both cheaper and better 2. Testers praised its animated visuals and agentic video editing 1.

But benchmarks score finished tasks; power users complained about staying on task through long delegated work and explaining what it did afterward, exactly the work a trillion-dollar premium is priced on 1.

The first print decides how much the gap is worth

Thursday's ANT debut prices the gap as it stands 36. The bear case for Anthropic is OpenAI's own parity claim landing a release or two from now; until users stop redoing Astra's work, the stumbles defend the premium. Watch whether the stock clears $80 on the first print, and whether OpenAI ships a fix faster than its usual cadence 6.

Get Model by email, free

The AI labs, their models, and the business behind them. In our daily email, every weekday.

Reader comments

0 comments

    Model

    HeyGen's Avatar 6 scored 84% on realism, but its 42.2% pick rate was below chance

    Avatar 6 launched November 17, 2026, with an 84% realism figure and first place in a blind arena; the score doubles a below-chance pick rate, and the arena capped 97% of clips at 35 seconds. That is the evidence base under a company claiming $200 million of annual revenue.

    Model

    MiniMax hit a record low, about 90% below its January peak, as a US rule gates its rented Nvidia compute

    Washington's new remote-access rule puts a US license between China's listed model makers and the offshore Nvidia capacity they rent. MiniMax, the purest listed bet on Chinese foundation models, holds the price tag.

    Model

    Microsoft books Azure revenue on Chinese open models running on Baseten hardware it does not own

    Microsoft now bills customers' committed Azure dollars for open models from Chinese labs running on Baseten's hardware, a toll booth earning rent on compute it does not own.

    Model

    Alibaba stops giving away its best AI model, three days before earnings

    Alibaba launched Qwen4-Max on November 16 and kept the 2.4-trillion-parameter flagship behind its own paid API, promising open weights only for smaller models. Thursday's report is the first test of whether Qwen's three billion downloads become cloud revenue, and it lands two days after a US rule put the chips behind the model under license.

    Model

    Perplexity claims $1 billion of revenue on five times less spend than Anthropic's $1.4 trillion IPO

    Perplexity says it hit $1 billion in recurring revenue on a fraction of Anthropic's spend, three days before Anthropic prices a $1.4 trillion IPO on the opposite assumption. Whose economics hold will only be settled by a filing Srinivas now says could come in early 2027.

    Model

    Anthropic prices a $1.4 trillion IPO as Databricks logs show its costliest model at 0.6% of enterprise tokens

    Databricks' gateway logs show the costliest Claude tier at 0.6% of enterprise tokens and open-weight models near a third, hours before Anthropic prices a $1.4 trillion debut on premium-model revenue.

    Subscribe to TheInference for free