Event · Real-world reviews

GPT-6.1 Terra, as reviewed by the people who ran it: '$1 in 10 minutes' vs. 'why would anyone use this?'

OpenAI released its mid-tier GPT-6.1 Terra on Oct 26. Per OpenAI, API pricing is $1 per million input tokens and $5 per million output — half of GPT-5.6 Terra and GPT-6.1 Sol — and at max effort it costs about 9x less per task than Claude Opus 5.5. OpenAI itself says Sol and Astra remain better for the hardest coding. Here's what people who tested it on X said (OpenAI staff excluded).

1. Mixed@notjazii · X ↗

   'Sonnet: 1.5 hours and $18. Terra: 10 minutes and $1'

   Developer notjazii ran the same prompt at max reasoning on Claude Sonnet 5.5 and Terra. Sonnet took 1 hour 30 minutes and $18; Terra took 10 minutes and $1, and the results "came out totally different."

2. Negative@atomic_chat_hq · X ↗

   'Sonnet crushed Terra'

   AI app maker Atomic Chat gave teams of eight agents the same game-building prompt. Terra rushed it in 6 minutes for $1.06; Sonnet 5.5 spent an hour and $18.10 on "a far better game."

3. Positive@MrTacticalX · X ↗

   Faster and cheaper than Haiku

   MrTacticalX had both build a three.js snow globe: Terra took 2 minutes and about $0.13, Claude Haiku 5.5 took 9 minutes and about $0.32. Against Sonnet 5.5: Terra 2 minutes and $0.13, Sonnet 35 minutes and $7.08.

4. Negative@vib3coded · X ↗

   'Opus dominates again'

   vib3coded asked both for a small courtyard scene: $14.17 for Opus 5.5 versus $1.23 for Terra. "Those rich colors, the lighting, all the little details — the whole courtyard feels alive," he wrote, siding with Opus.

5. Mixed@intheworldofai · X ↗

   Sol, at 4x the price, added a working time slider

   intheworldofai tried an animated NYC SVG: Terra spent $0.12, 4m32s and 25K tokens on a "polished postcard-style illustration"; GPT-6.1 Sol spent $0.47, 8m20s and 49K tokens on an interactive cityscape with a day-to-night slider.

6. Negative@kimmonismus · X ↗

   'Why would I use it when Sonnet is much better?'

   AI news account kimmonismus wrote: "At least they are honest with the benchmarks. I get it, it's cheap. But why would I use it when sonnet 5.5 is much better and even 6.1 Sol is almost the same pricing but better?" calling it "the weirdest release" (800 likes).

7. Positive@davis7 · X ↗

   'Not for us — for the $20/month subscribers'

   davis7: "If ur reading this post u will probably never select Terra for ur work unless ur building an agent into a product," but "this is however huge for the $20/mo. subs and normal people just using chatgpt" (400 likes).

8. Negative@Mr_Salio · X ↗

   'Only crazy speed'

   After comparing it with Opus 5.5, Mr_Salio called Terra "one of the most disappointing models I've tested since GPT-5.5" with "only a crazy speed," adding OpenAI "would've been better off keeping this model in the office."

9. Positive@ArtificialAnlys · X ↗

   Independent benchmark: 'pushing the cost efficiency frontier'

   Artificial Analysis said Terra is "more intelligent for less than a third of the cost" of GPT-5.6 Terra, pushing the cost-efficiency frontier (1,400 likes).