OpenAIReasoning modelNewReleased 22 Sept 2026

GPT-6 Sol

Reasoning effort

GPT-6 Sol is an OpenAI proprietary reasoning model, released 22 Sept 2026. Its best configuration (max reasoning effort) ranks #22 of 432 on the BenchLeader Index at 67.7 ±9.5, in the top 25. It scores highest in reasoning (95) and lowest in multimodal (67). At $4.00 per million tokens blended it is among the most expensive ranked models. Output speed of 126 tokens per second puts it faster than most, with a first answer in 107.2 s. It has been measured at 6 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 22 Sept 2026.

Blended price: $4.00/M; $2.00 in · $10.00 out
Output speed: 126 tok/s; measured by Artificial Analysis
First answer: 107 s; first token 2.77 s
Context: 1.1M
Full answer: 111 s; median, reasoning included
Cost per run: $1.06; one full Intelligence Index run
Released: 22 Sept 2026

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index68
Reasoning95
Knowledge75
Multimodal67
Long context68
Composite86

Versions

OpenAI has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

Model	Released	Index	Rank
NewGPT-6 Solmaxthis page	22 Sept 2026	67.7	#22
GPT-5.6 Solmax	9 Jul 2026	68.4	#19

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
no reasoning	54.1	#249	116 tok/s	0.97 s	$0.0038	Composite 63 · Knowledge 63 · Long context 58 · Multimodal 52 · Reasoning 49
low	61.7	#82	124 tok/s	1.44 s	$0.0038	Composite 70 · Knowledge 75 · Long context 65 · Multimodal 63 · Reasoning 71
medium	64.9	#43	114 tok/s	1.96 s	$0.0038	Composite 77 · Knowledge 75 · Long context 67 · Multimodal 64 · Reasoning 85
high	65.7	#36	123 tok/s	7.57 s	$0.0038	Composite 81 · Knowledge 75 · Long context 68 · Multimodal 65 · Reasoning 87
xhigh	66.4	#33	134 tok/s	45 s	$0.0038	Composite 82 · Knowledge 75 · Long context 67 · Multimodal 66 · Reasoning 91
maxbest	67.7	#22	126 tok/s	107 s	$0.0038	Composite 86 · Knowledge 75 · Long context 68 · Multimodal 67 · Reasoning 95

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	no reasoning	low	medium	high	xhigh	max	Source
Humanity's Last Exam (AA)not in index	18.4%#200	34.9%#98	41.0%#63	44.1%#39	46.3%#33	47.9%#25	Artificial Analysis
CritPt	4.0%#138	16.3%#68	24.6%#36	25.4%#33	28.0%#21	30.9%#7	Artificial Analysis

Coding

Benchmark	no reasoning	low	medium	high	xhigh	max	Source
SciCode (AA)not in index	47.3%#102	50.2%#86	53.8%#59	54.9%#48	55.1%#43	57.6%#21	Artificial Analysis
Terminal-Bench 4.0 (AA)not in index	13.1%#51	9.1%#64	18.7%#43	26.3%#32	30.3%#30	43.9%#16	Artificial Analysis

Agents & tools

Benchmark	no reasoning	low	medium	high	xhigh	max	Source
GDPval (AA)not in index	36.3%#87	33.8%#97	41.0%#70	43.8%#56	46.8%#46	49.4%#35	Artificial Analysis
AutomationBenchnot in index	–	–	–	–	–	61.6%#14	Artificial Analysis
GDP.pdfnot in index	–	–	–	–	–	24.8%#15	Artificial Analysis
AA-Briefcasenot in index	–	–	–	–	–	1483#26	Artificial Analysis

Knowledge

Benchmark	no reasoning	low	medium	high	xhigh	max	Source
AA-Omniscience	-0.8#116	26.5#38	27.0#35	26.8#36	26.7#37	27.1#34	Artificial Analysis
AA-Omniscience: accuracynot in index	45.2%#76	51.3%#49	53.5%#40	53.7%#38	53.9%#37	54.5%#35	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	16.0%#299	49.3%#102	43.2%#124	41.9%#127	41.1%#130	39.9%#135	Artificial Analysis

Multimodal

Benchmark	no reasoning	low	medium	high	xhigh	max	Source
MMMU-Pro	68.2%#152	78.8%#56	80.6%#38	81.2%#33	82.4%#26	83.3%#22	Artificial Analysis

Long context

Benchmark	no reasoning	low	medium	high	xhigh	max	Source
AA-LCR	64.0%#228	79.3%#81	82.3%#31	83.7%#14	81.3%#45	83.7%#14	Artificial Analysis
MLCRnot in index	–	–	–	–	–	16.1%#24	Artificial Analysis

Composite

Benchmark	no reasoning	low	medium	high	xhigh	max	Source
AA Intelligence Index	28.1#111	33.9#74	39.8#46	42.8#36	44.1#31	47.5#18	Artificial Analysis

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
Amazon Bedrock	88 tok/s	1.96 s	$2.20	$11.00	1.1M	–
Azure	73 tok/s	4.94 s	$2.20	$11.00	1.1M	–
OpenAI	67 tok/s	3.03 s	$1.00	$5.00	1.1M	–
OpenAI	53 tok/s	2.77 s	$2.00	$10.00	1.1M	–
OpenAI	38 tok/s	3.22 s	$4.00	$20.00	1.1M	–
Azure	37 tok/s	4.99 s	$2.00	$10.00	1.1M	–
Azure	12 tok/s	1.73 s	$2.20	$11.00	1.1M	–

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.200 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0038	$0.0033	1.8 min
Summarise a 30-page report	12,000 / 600	$0.030	$0.014	1.9 min
Code edit	6,000 / 1,500	$0.027	$0.019	2.0 min
Agentic coding session	60,000 / 4,000	$0.160	$0.079	2.3 min
Structured extraction	2,000 / 200	$0.0060	$0.0033	1.8 min