API ManyНовые модели: Qwen 3.8 Max, Claude Haiku 5.5, Claude Opus 5.5, Claude Sonnet 5.5
OpenAILLMComparison

GPT-6.1 Terra vs GPT-6.1 Sol: Price, Benchmarks and Which to Use

GPT-6.1 Terra costs half as much as GPT-6.1 Sol ($1 / $5 vs $2 / $10 per 1M tokens); Sol scores higher (55 vs 51 on the AA Intelligence Index at max). Specs, costs and when to pick each.

29 окт. 2026 г. · API Many

GPT-6.1 Sol (September 29, 2026) and GPT-6.1 Terra (October 26, 2026) share the same context window, output limit and tool support. The differences are price, peak capability and speed: Terra costs half as much per token, while Sol scores higher and is the model OpenAI recommends for the most complex coding.

Sources:OpenAI: GPT-6.1 Sol system card addendum (Sep 29, 2026) · OpenAI: GPT-6.1 Terra system card addendum (Oct 26, 2026) · OpenAI: Introducing GPT-6.1 Terra (Oct 26, 2026)

Specs and prices

GPT-6.1 TerraGPT-6.1 Sol
ReleasedOctober 26, 2026September 29, 2026
Input / outputText and image in; text outText and image in; text out
Context window1,050,000 tokens1,050,000 tokens
Maximum output128,000 tokens128,000 tokens
Knowledge cutoffJune 1, 2026April 30, 2026
Reasoning effortnone to maxlow to max
Input / output per 1M tokens$1 / $5$2 / $10
Cached input per 1M tokens$0.05$0.10
Cache write (30-minute) per 1M tokens$1.25$2.50
Over 272K input tokens (input / output)$2 / $7.50$4 / $15
Batch (input / output)$0.50 / $2.50$1 / $5
Faster tiersFast mode (2x)Fast mode (2x), Ultrafast (6x)
On API ManyNoYes

Sources:OpenAI API docs: GPT-6.1 Terra model page · OpenAI API docs: GPT-6.1 Sol model page · OpenAI API pricing

Independent scores by effort level

Artificial AnalysisGPT-6.1 TerraGPT-6.1 Sol
Intelligence Index v5, max effort51 ($0.96 per task)55 ($1.27 per task)
Intelligence Index v5, xhigh45 ($0.40)52 ($0.76)
Intelligence Index v5, high42 ($0.26)50 ($0.59)
Intelligence Index v5, medium38 ($0.17)46 ($0.36)
Coding Agent Index v2.04854
Output speed (max effort)91 tokens/s81 tokens/s
Index scores, cost per Index task and speed as listed on October 29, 2026; the Coding Agent Index figures are from Artificial Analysis’s Terra launch article.

Sources:Artificial Analysis: LLM leaderboard · Artificial Analysis: GPT-6.1 Terra release page · Artificial Analysis: GPT-6.1 Terra pushes the cost efficiency frontier (Oct 26, 2026)

At every effort level Sol scores higher, by 4 to 8 points, and costs roughly 1.3x to 2.3x as much per task. Terra at max effort (51, $0.96) scores about the same as Sol at high effort (50, $0.59), and Sol at high costs less per task.

Source:Artificial Analysis: LLM leaderboard

What OpenAI says about the two

  • Speed: on Agents’ Last Exam, Terra completes tasks over three times faster than GPT-6.1 Sol and GPT-6 Astra.
  • Science: on Terminal-Bench-Science 0.1 at max effort, Terra surpasses Sol at lower cost.
  • Visual reasoning: on MMMU-Pro, Terra nearly matches Sol and GPT-6 Astra at much lower cost.
  • Coding: for the most complex coding tasks, OpenAI says GPT-6.1 Sol and GPT-6 Astra remain the better choice.
  • Factuality: in the Terra system card, Terra makes substantially fewer factual errors than GPT-6 Sol, on par with GPT-6.1 Sol.

Sources:OpenAI: Introducing GPT-6.1 Terra (Oct 26, 2026) · OpenAI: GPT-6.1 Terra system card addendum (Oct 26, 2026)

From the Terra system card

OpenAI system cardGPT-6.1 TerraGPT-6.1 Sol
ExploitBench internal port: arbitrary code execution24.5%21.5%
Sandbox Bench: targets exploited5 of 229 of 22
HealthBench Professional (length-adjusted)62.764.2
MentalHealthBench (overall)55.5%57.9%
Fails to say the search tool is broken1.83% of cases2.08% of cases
Reported by OpenAI in the GPT-6.1 Terra system card addendum, mostly as safety and preparedness evaluations.

Source:OpenAI: GPT-6.1 Terra system card addendum (Oct 26, 2026)

Which to pick

  • High-volume work such as extraction, form filling, screen reading and drafting, where cost per call matters most: Terra, at half Sol’s token price.
  • Hard coding, long agent runs and research tasks: Sol. It scores higher at every effort level, and OpenAI points complex coding to Sol or Astra.
  • Tight latency budgets: Terra, which OpenAI reports finishes tasks over three times faster.
  • You want to run it on API Many today: Sol. Terra is not currently offered here.

On API Many, GPT-6.1 Sol is available from $2.00 вход · $10.00 выход. GPT-5.6 Terra, the previous Terra, is also available, from $2.00 вход · $12.00 выход. Note that on API Many it is not cheaper than GPT-6.1 Sol: input costs the same and output costs more.

Related

FAQ

Is GPT-6.1 Terra cheaper than GPT-6.1 Sol?

Yes. Per token, Terra costs half as much: $1 / $5 per million input / output tokens versus $2 / $10. On the Artificial Analysis Intelligence Index, Terra also costs less per task at each effort level, but Sol at high effort scores about the same as Terra at max for less.

Which is smarter, GPT-6.1 Terra or GPT-6.1 Sol?

Sol. Artificial Analysis scores it 55 at max effort versus 51 for Terra, and 54 versus 48 on its Coding Agent Index.

Do they have the same context window?

Yes: 1,050,000 tokens, up to 922,000 input tokens and 128,000 output tokens. Terra’s knowledge cutoff is later (June 1, 2026 vs April 30, 2026).

Which one can I use on API Many?

GPT-6.1 Sol. GPT-6.1 Terra is not currently offered on API Many.

Sources

  1. OpenAI API docs: GPT-6.1 Terra model page — GPT-6.1 Terra specs and pricing notes.
  2. OpenAI API docs: GPT-6.1 Sol model page — GPT-6.1 Sol specs, reasoning efforts and pricing notes (including Ultrafast).
  3. OpenAI API pricing — Standard, long-context and Batch prices for both models.
  4. OpenAI: Introducing GPT-6.1 Terra (Oct 26, 2026) — OpenAI’s speed, science, visual reasoning and coding guidance for Terra versus Sol.
  5. OpenAI: Introducing GPT-6.1 Sol — GPT-6.1 Sol positioning and prices.
  6. OpenAI: GPT-6.1 Terra system card addendum (Oct 26, 2026) — Terra release date and system card evaluations.
  7. OpenAI: GPT-6.1 Sol system card addendum (Sep 29, 2026) — GPT-6.1 Sol release date.
  8. Artificial Analysis: LLM leaderboard — Intelligence Index v5 scores, cost per task and output speed by effort level.
  9. Artificial Analysis: GPT-6.1 Terra pushes the cost efficiency frontier (Oct 26, 2026) — Coding Agent Index v2.0 scores.
Попробуйте с одним API-ключом

Все модели из статьи доступны в API Many. Оплата по факту от $10.

Смотреть модели →

Ещё в блоге