GPT-6.1 Sol: Benchmarks, Rankings and Price

OpenAI's mid-priced flagship, described by OpenAI as near-Astra performance at a lower cost. It scores close to GPT-6 Astra on most tests while costing less than a quarter as much per Artificial Analysis task.

Where people rank it

Reading images and documents (tied), 1,813 votes

Coding and building apps (tied), 2,863 votes

Web search and research (tied), 5,555 votes

From blind votes on LMArena, where people pick the better of two anonymous answers. See the full tables on the AI leaderboard.

On paper

Fixed-test rankings from the benchmarks tab: Value #4, Reasoning #6, Speed #17.

Strengths and weaknesses

Artificial Analysis Intelligence Index of 51.8 at $0.72 per task, the cheapest of the models here that score above 50
94.2% on ARC-AGI-2 and a top-four score on the LMArena Code Arena
In the tied top group of the LMArena Vision leaderboard

Slow output on Artificial Analysis (66 tokens per second)
Mid-table on the LMArena text and creative writing leaderboards

Best for

Coding with OpenAI Codex
A strong all-rounder at a mid price
Reading charts, screenshots and documents

Every score we have

Each bar compares this model with the best result any model on this page has on the same test.

LMArena text arena (human votes)1,483.8Arena score 1483.8 (95% CI 1475.7 to 1492.0); 6,320 votes; entry "gpt-6.1-sol-max"; position 22 of 416 on the full board
LMArena creative writing1,458.4Arena score 1458.4 (95% CI 1441.5 to 1475.4); 1,419 votes; entry "gpt-6.1-sol-max"; position 30 of 414 on the full board
LMArena WebDev (coding)1,755.5Arena score 1755.5 (95% CI 1742.0 to 1768.9); 2,863 votes; entry "GPT-6.1 Sol (Max)"; position 4 of 144 on the full board
LMArena vision1,294.1Arena score 1294.1 (95% CI 1279.6 to 1308.5); 1,813 votes; entry "gpt-6.1-sol-max"; position 8 of 165 on the full board
LMArena Agent Arena0.136IPS score 0.1360 (95% CI 0.1170 to 0.1551); 13,444 sessions; entry "GPT-6.1 Sol (Max)"; position 4 of 52 on the full board
Artificial Analysis Intelligence Index51.851.8 on the Artificial Analysis Intelligence Index v4.3.2 (AA shows 52), setting "GPT-6.1 Sol (max)"
Humanity's Last Exam (Artificial Analysis)52.9%52.9% on Humanity's Last Exam (AA, 2,158 text-only questions), setting "GPT-6.1 Sol (max)"
Terminal-Bench 4.0 (Artificial Analysis run)56.1%56.1% on Terminal-Bench 4.0 as run by AA (mini-swe-agent harness), setting "GPT-6.1 Sol (max)"
Long-context reasoning (AA-LCR)83.0%83.0% on AA-LCR (long-context reasoning), setting "GPT-6.1 Sol (max)"
MMMU-Pro, image understanding86.0%86.0% on MMMU-Pro (image understanding, AA), setting "GPT-6.1 Sol (max)"
AA-Omniscience, accuracy minus hallucination41.541.5 on the AA-Omniscience Index (knowledge accuracy minus hallucination, range -100 to 100), setting "GPT-6.1 Sol (max)"
Output speed, tokens per second66 t/s66 output tokens per second (AA median), row "GPT-6.1 Sol (max)"
Cost per Artificial Analysis test task$0.72$0.72 to run one AA Intelligence Index task, row "GPT-6.1 Sol (max)"
Artificial Analysis Coding Agent Index62.962.9 on the AA Coding Agent Index v1.5 (DeepSWE v1.1, Terminal-Bench 4.0, SWE-Atlas-QnA), best listed setup "Codex - GPT-6.1 Sol (xhigh)"
Terminal-Bench 4.0 leaderboard58.2%58.2% resolved on Terminal-Bench 4.0 (plus or minus 3.1, 95% CI) with Codex at max effort
ARC-AGI-2 puzzles94.2%94.17% on ARC-AGI-2 (semi-private set), best listed setting "GPT-6.1 Sol (Max)", $0.25 per task
ARC-AGI-3 puzzles52.7%52.73% on ARC-AGI-3 (semi-private, standard harness; "provider adapter" runs excluded), best listed setting "GPT-6.1 Sol (Max)"

GPT-6.1 Sol

Where people rank it

On paper

Strengths and weaknesses

Best for

Every score we have

Compare with