$1 a Million Tokens. OpenAI Just Priced Itself to Commodity.
GPT-6.1 Terra undercuts OpenAI's own month-old flagship by half. The pattern points to a lab that has stopped selling intelligence and started racing to own the agent doing your job.

Every AI outlet, every Hacker News thread ran the same story Monday, Oct. 26: OpenAI shipped a new model. That is not news. Labs ship models every few weeks. What they don't usually do is make the new one cheaper than the one they released a month earlier.
GPT-6.1 Terra is not expensive enough to be the premium release the coverage treats it as.
The verdict: OpenAI didn't ship a smarter model this week. It shipped a cheaper one, and the gap between those two things is the entire story.
The Price Did the Talking
Standard pricing is $1 per million input tokens and $5 per million output tokens. That's half the price of Claude Sonnet 5.5 and GPT-5.6 Terra, and half the price of GPT-6.1 Sol — the model OpenAI was selling as its mid-tier flagship weeks earlier. Terra shipped at half Sol's price four weeks after Sol's own launch.
A lab does not cut its own newest price in half unless it has decided the margin on tokens matters less than who owns the workflow built on top of them.
Line the whole OpenAI stack up: GPT-6 Astra runs $10 input / $50 output, GPT-6.1 Sol $2 / $10, GPT-6.1 Terra $1 / $5, and GPT-6 Luna $0.10 / $0.50. That's not a product ladder. That's a company racing itself to the floor on the one tier built specifically for machines that act.
| Input (tokens) | |
|---|---|
| GPT-6 Astra | $10/M |
| GPT-6.1 Sol | $2/M |
| GPT-6.1 Terra | $1/M |
| GPT-6 Luna | $0.1/M |

Translation: We Stopped Selling You Brains
OpenAI measured GPT-6.1 Terra's computer-use capabilities on Agents' Last Exam, a benchmark of professional tasks that includes turning architectural drawings into a 3D model and reviewing film shot by shot. Translation: we are measuring how well it replaces a junior employee's afternoon, not how well it chats.
OpenAI says Terra performs on par with GPT-6 Astra at only one-tenth its price per token. Against Anthropic's top model, Terra is over three times as token efficient as Claude Opus 5.5 and one-fourth its price per token, resulting in roughly 9x lower cost per task. Independent numbers from Artificial Analysis put the task cost at $1.01 for Terra against Claude Opus 5.5's $8.82 on the same exam.
Simon Willison, who has watched every OpenAI release since GPT-3, called it correctly: Terra sits in the same position as Sonnet in the Anthropic collection — not the cheapest or most expensive, making it harder to decide when to use it. For subscribers paying ChatGPT subscriptions rather than API bills, he was blunt: for subscribers, it's not a bombshell release.
That line matters.
If Terra isn't a bombshell for the chat product, the chat product was never the point. The API pricing was.
Who Just Got Squeezed
A lab racing to win a capability contest doesn't discount its newest model against its own month-old release. A lab racing to own the agent layer does exactly that, because token price is the only lever that decides whether enterprises route high-volume, repetitive computer-use work through you or through the next guy.
Terra already costs half as much per token as Claude Sonnet 5.5 and Gemini 4 Argon's introductory price. One analysis named the collateral damage directly: the launch targets Anthropic's mid-tier story and anyone selling cheaper Sonnet-class pricing.
If you built a sales pitch around "frontier intelligence costs more, budget accordingly," that pitch just lost its floor.
What to Watch
Watch whether Anthropic and Google answer this price cut within days rather than weeks — a slow response is itself the signal they're not ready to compete on agent economics. Watch whether enterprise capex for Q4 2026 shifts from "evaluate AI" line items to "deploy agents" line items now that the per-task math works. And watch whether OpenAI's next release undercuts Terra, too — because a lab racing to own a layer doesn't stop at one price cut.
- Introducing GPT-6.1 Terra
- GPT-6.1 Terra
- OpenAI Launches GPT-6.1 Terra, Halving GPT-6.1 Sol's API Price for Agent Work
- GPT-6.1 Terra halves OpenAI's API prices and still gains ground on its larger models
- GPT-6.1 Terra Arrives at Half the Price, With 37 Fewer Error Points and No Flagship in Sight
- GPT-6.1 Terra (Hacker News discussion)
- GPT-6.1 Terra vs Claude Sonnet 5.5, Gemini 4 Argon, Grok 4.7: API pricing compared
- GPT-6.1 Terra: $1/$5 Pricing, Benchmarks, and When to Pick It
- GPT-6.1 Terra: Benchmarks, Pricing and Access
- GPT-6.1 Terra: Benchmarks, Pricing, Specs, and How It Compares
- OpenAI launches GPT-6.1 Terra with 3x faster AI agents, 50% lower pricing
- OpenAI: GPT-6.1 Terra - API Pricing & Benchmarks
- GPT 6.1 Terra: Features, Pricing, and Access (2026)
- OpenAI Launches GPT-6.1 Terra With $1/M Input Pricing
- GPT-6.1 Terra: $1/$5 Pricing, Benchmarks, Computer Use
- Simon Willison’s Weblog
- Simon Willison on X: "Got access to GPT-6 Astra. Want to see some pelicans? Yeah you want to see some pelicans... here's a grid comparing Astra to GPT-5.6 Sol, Terra, and Luna https://t.co/LJi3XOpYpJ" / X
- GPT-6.1 Terra Benchmarks & Pricing (October 2026)
- Simon Willison on gpt-6-astra
- Willison benchmarks GPT-6 Astra pelicans vs GPT-5.6 tiers
- GPT-6.1 Terra pushes the cost efficiency frontier
- What Is GPT-6.1 Terra? The 18-Config Judge Model
- GPT-6.1 Terra Launch: Benchmarks, Pricing
- OpenAI's GPT-6.1 Terra Cuts Agent Costs 50% and Runs 3x Faster
- GPT-6.1 Terra Review: Capabilities, Uses and API Pricing
- GPT-6.1 Terra Officially Released: Rock-Bottom Pricing Unveils the Definitive Working Model for the Agent Era
- OpenAI Releases GPT-6.1 Terra: API Prices Halved, Agent Task Costs Drop to One-Tenth of Flagship Model — BigGo Finance