What are GPT-5.6 Terra's benchmark scores?

As of 24 Oct 2026, GPT-5.6 Terra (max effort) scores 64.4 ±5.0 on the BenchLeader Index, top 100, strongest in composite (77) and weakest in knowledge (57). That rests on 22 independent benchmark results, and every score on this page links to its publisher.

How does GPT-5.6 Terra rank?

GPT-5.6 Terra (max effort) is #51 of 785 ranked configurations on 24 Oct 2026 (top 100). 142 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.

How much does GPT-5.6 Terra cost per 1M tokens?

As of 24 Oct 2026, GPT-5.6 Terra costs $2.00 per 1M input tokens and $12.00 per 1M output tokens, $4.50 blended (three input tokens per output token).

Is GPT-5.6 Terra better than Gemini 3.8 Flash?

As of 24 Oct 2026, Gemini 3.8 Flash (high) and GPT-5.6 Terra (max) are neck and neck on quality (64.4 vs 64.4 on the BenchLeader Index, inside the ±6.3 margin of error), and Gemini 3.8 Flash (high) costs 3.0× less, which makes it the better buy for most workloads. The head-to-head page has the benchmark-by-benchmark detail.

When was GPT-5.6 Terra released?

OpenAI released GPT-5.6 Terra on 9 Jul 2026. BenchLeader has results for it at 7 reasoning-effort settings.

OpenAIReasoning modelCompare with another model

GPT-5.6 Terra

Reasoning effort

GPT-5.6 Terra is an OpenAI proprietary reasoning model, released 9 Jul 2026. Its best configuration (max reasoning effort) ranks #51 of 785 on the BenchLeader Index at 64.4 ±5.0, in the upper half. The ± is the point: 142 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in composite (77) and lowest in knowledge (57). At $4.50 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 91.6 s. It has been measured at 7 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 20 Oct 2026.

Blended price: $4.50/M; $2.00 in · $12.00 out
Output speed: 126 tok/s; measured by Artificial Analysis
First answer: 92 s; first token 1.83 s
Context: 1.1M
Full answer: 96 s; median, reasoning included
Cost per run: $1.40; one full Intelligence Index run
Released: 9 Jul 2026

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

Overall index64
Reasoning70
Coding65
Agents & tools59
Maths69
Knowledge57
Instruction following70
Multimodal64
Long context67
Composite77

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	49.5	#393	115 tok/s	0.73 s	$0.0044	Agents & tools 44 · Coding 47 · Composite 53 · Knowledge 51 · Long context 54 · Maths 47 · Multimodal 50 · Reasoning 50
low	56.5	#200	114 tok/s	1.39 s	$0.0044	Agents & tools 55 · Coding 54 · Composite 60 · Instruction following 60 · Knowledge 58 · Long context 61 · Maths 62 · Multimodal 59 · Reasoning 51
medium	56.9	#193	109 tok/s	1.63 s	$0.0044	Agents & tools 54 · Coding 55 · Composite 63 · Instruction following 62 · Knowledge 59 · Long context 62 · Multimodal 60 · Reasoning 56
high	61.3	#97	107 tok/s	2.15 s	$0.0044	Agents & tools 66 · Coding 63 · Composite 68 · Instruction following 64 · Knowledge 60 · Long context 64 · Multimodal 62 · Reasoning 64
xhigh	62.8	#76	112 tok/s	7.49 s	$0.0044	Agents & tools 64 · Coding 60 · Composite 72 · Human preference 66 · Instruction following 66 · Knowledge 60 · Long context 65 · Maths 66 · Multimodal 62 · Reasoning 64
maxbest	64.4	#51	126 tok/s	92 s	$0.0044	Agents & tools 59 · Coding 65 · Composite 77 · Instruction following 70 · Knowledge 57 · Long context 67 · Maths 69 · Multimodal 64 · Reasoning 70
not stated	52.7	#307	65 tok/s	1.83 s	$0.0044	Coding 55 · Composite 59 · Human preference 55 · Reasoning 49

What more thinking costs

Turning the effort up buys index points and multiplies the bill. Here max costs 10× no reasoning for +14.9 on the index — 4.5 points per doubling of spend. Compare that with other models.

How its index has moved

8 Sept 2026 to 23 Oct 2026

max
xhigh
high
medium

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. 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	not stated	Source
GPQA Diamond	77.3%#136	87.4%#73	–	–	–	93.3%#18	–	Epoch AI Benchmarking Hub
SimpleBench	–	–	–	–	48.9%#45	–	48.9%#45	SimpleBench
Arena Hard Prompts	–	–	–	–	1490#48	–	–	Arena (LMArena)
LiveBench 2.0 Reasoningnot in indexeffort not stated	–	–	–	–	–	–	90.6%#12	LiveBench
LiveBench 1.0 Reasoningnot in index	–	–	–	–	–	90.6%#11	–	LiveBench
GPQA Diamond (AA)not in index	74.7%#258	84.3%#143	87.2%#102	89.6%#71	90.8%#55	92.5%#35	–	Artificial Analysis
Humanity's Last Exam (AA)not in index	11.4%#294	29.2%#158	33.3%#139	38.5%#98	41.9%#77	42.9%#63	–	Artificial Analysis
GPQA Diamond (Vals)not in index	–	–	–	–	90.9%#24	–	–	Vals AI
Kagi LLM Benchmarkeffort not stated	–	–	–	–	–	–	51.3%#82	Kagi LLM Benchmark
ARC-AGI-1	–	60.2%#161	77.0%#127	92.0%#60	94.0%#46	96.5%#23	–	ARC Prize
ARC-AGI-2	–	18.8%#145	37.5%#122	67.1%#69	74.2%#50	83.9%#35	–	ARC Prize
ARC-AGI-3	–	0.0%#67	0.1%#64	0.5%#46	0.7%#43	0.8%#41	–	ARC Prize
CritPt	2.0%#201	9.4%#124	17.4%#79	22.9%#54	27.1%#33	30.0%#16	–	Artificial Analysis
Chess Puzzlesnot in index	5.0%#132	27.0%#72	–	–	–	47.0%#24	–	Epoch AI Benchmarking Hub
Mystery Game Puzzlesnot in index	14.0%#97	19.0%#64	10.0%#112	–	–	35.0%#23	–	Epoch AI Benchmarking Hub
LMCAnot in index	46.5%#83	49.3%#70	49.6%#68	51.4%#59	52.2%#55	55.0%#46	–	Epoch AI Benchmarking Hub
DTBenchnot in index	78.7%#138	90.1%#82	90.1%#82	91.2%#73	92.5%#68	93.3%#64	–	Epoch AI Benchmarking Hub

Coding

Benchmark	no reasoning	low	medium	high	xhigh	max	not stated	Source
SciCode	45.1%#139	49.9%#111	50.5%#103	52.4%#86	52.3%#87	55.0%#52	–	SciCode
WeirdML	–	–	–	78.3%#20	–	–	–	WeirdML
FrontierCodeeffort not stated	–	–	–	–	–	–	41.3%#24	Cognition
Arena Coding	–	–	–	–	1518#42	–	–	Arena (LMArena)
Arena WebDev	–	–	–	–	1521#52	–	–	Arena (LMArena)
LiveBench 2.0 Codingnot in indexeffort not stated	–	–	–	–	–	–	86.2%#19	LiveBench
LiveBench 1.0 Codingnot in index	–	–	–	–	–	78.3%#31	–	LiveBench
SciCode (AA)not in index	45.1%#146	49.9%#113	50.5%#105	52.4%#86	52.3%#87	55.0%#56	–	Artificial Analysis
LiveCodeBench	–	–	–	–	85.9%#25	–	–	Vals AI
IOI	–	–	–	–	–	87.6%#7	–	Vals AI
SWE-bench (Vals)not in index	–	–	–	–	–	95.4%#5	–	Vals AI
Code Migrationnot in index	–	–	–	–	–	47.8%#13	–	Vals AI
ProgramBenchnot in index	–	–	–	–	–	0.5%#18	–	Vals AI
Vibe Code Bench 1-100not in index	–	–	–	–	–	14.8%#19	–	Vals AI
Vibe Code Bench v1.1not in index	–	–	–	–	–	74.6%#37	–	Vals AI
Surface Evolver Benchnot in index	–	–	–	–	83.8%#7	–	–	Epoch AI Benchmarking Hub
DeepSWE v1.1not in index	–	24.1%#66	35.1%#61	53.8%#42	60.2%#32	69.6%#14	–	Epoch AI Benchmarking Hub
CursorBenchnot in index	–	25.2%#73	27.6%#71	30.7%#66	33.6%#54	41.3%#37	–	Cursor
ALE-Benchnot in index	–	–	–	–	–	1951.4#10	–	Epoch AI Benchmarking Hub
Terminal-Bench 4.0 (AA)not in index	0.5%#139	1.5%#122	1.0%#128	1.5%#122	10.1%#89	35.4%#34	–	Artificial Analysis
Terminal-Bench 2.1 (AA)not in index	56.2%#105	62.5%#89	72.3%#69	75.7%#61	80.2%#44	88.0%#12	–	Artificial Analysis
Startrise Frontendnot in indexeffort not stated	–	–	–	–	–	–	66.9%#22	Startrise

Agents & tools

Benchmark	no reasoning	low	medium	high	xhigh	max	not stated	Source
Terminal-Bench	–	–	–	–	–	21.5%#70	–	Terminal-Bench
APEX-Agents	–	–	–	–	–	58.2%#15	–	Mercor
Arena Agent	–	–	–	–	0.5#28	–	–	Arena (LMArena)
LiveBench Long Agenticnot in indexeffort not stated	–	–	–	–	–	–	48.4%#16	LiveBench
LiveBench Browsernot in indexeffort not stated	–	–	–	–	–	–	84.0%#10	LiveBench
LiveBench Agentic Codingnot in index	–	–	–	–	–	55.0%#30	–	LiveBench
Terminal-Bench Hard	–	43.9%#37	–	57.6%#11	62.9%#2	57.6%#11	–	Artificial Analysis
τ²-Bench Telecom (AA)not in index	–	60.5%#178	72.8%#147	78.4%#130	80.4%#123	86.3%#88	–	Artificial Analysis
Terminal-Bench 2.1 (Vals)	–	–	–	–	–	77.5%#14	–	Vals AI
GDPval-AA v2.1not in index	29.9%#86	30.5%#81	38.3%#66	43.9%#54	47.4%#44	47.6%#41	–	Artificial Analysis
τ³-Banking (AA)	15.7%#114	18.8%#102	25.6%#80	28.7%#72	29.7%#69	40.2%#29	–	Artificial Analysis
ITBench SRE (AA)not in index	–	–	–	–	–	51.0%#5	–	Artificial Analysis
APEX-Agents (AA)not in index	–	–	–	–	–	38.9%#4	–	Artificial Analysis
CyberBenchnot in index	–	–	–	–	–	72.1%#17	–	Vals AI
Finance Agent v2not in index	–	–	–	–	–	54.4%#27	–	Vals AI
Harvey's Legal Agent Benchmark v1not in index	–	–	–	–	–	0.8%#56	–	Vals AI
Legal Research Benchnot in index	–	–	–	–	–	41.4%#26	–	Vals AI
SkillsBenchnot in index	–	–	–	–	–	58.9%#11	–	Vals AI
SREBenchnot in index	–	–	–	–	–	9.5%#14	–	Vals AI
Tax Agent Benchnot in index	–	–	–	–	–	30.0%#28	–	Vals AI
Terminal-Bench 4.0 (Vals)not in index	–	–	–	–	–	22.7%#27	–	Vals AI
Terminal-Bench Sciencenot in index	–	–	–	–	–	10.0%#15	–	Vals AI
BALROGnot in index	–	–	–	–	–	53.2%#6	–	BALROG
Vending-Bench 2not in indexeffort not stated	–	–	–	–	–	–	7343.2#16	Andon Labs
AutomationBenchnot in index	12.5%#150	29.1%#117	33.5%#114	42.0%#95	47.1%#89	59.6%#39	–	Artificial Analysis
GDP.pdf (AA)not in index	11.2%#120	17.6%#80	17.0%#87	20.8%#59	24.6%#32	24.0%#38	–	Artificial Analysis
EnterpriseOps-Gymnot in index	–	–	–	–	–	38.5%#32	–	Artificial Analysis
Terminal-Bench-Science 0.1 (AA)not in index	–	–	–	–	–	13.8%#26	–	Artificial Analysis
FACTS Searchnot in indexeffort not stated	–	–	–	–	–	–	81.7%#9	Google

Maths

Benchmark	no reasoning	low	xhigh	max	Source
FrontierMath Tiers 1–3	–	–	–	86.0%#10	Epoch AI Benchmarking Hub
FrontierMath Tier 4	–	–	–	70.7%#20	Epoch AI Benchmarking Hub
OTIS Mock AIME	53.3%#178	88.9%#76	–	99.7%#13	Epoch AI Benchmarking Hub
ProofBench	–	–	74.0%#15	–	Vals AI
LiveBench Mathematicsnot in index	–	–	–	94.9%#18	LiveBench
Arena Maths	–	–	1483#36	–	Arena (LMArena)

Knowledge

Benchmark	no reasoning	low	medium	high	xhigh	max	not stated	Source
SimpleQA Verified	–	–	–	–	–	43.2%#41	–	Epoch AI Benchmarking Hub
LiveBench 2.0 Data Analysisnot in indexeffort not stated	–	–	–	–	–	–	79.3%#14	LiveBench
LiveBench Researchnot in indexeffort not stated	–	–	–	–	–	–	60.4%#43	LiveBench
LiveBench 1.0 Data Analysisnot in index	–	–	–	–	–	79.3%#15	–	LiveBench
AA-Omniscience	-23.2#256	-6.8#172	-5.1#164	-3.5#152	-3.0#151	0.1#137	–	Artificial Analysis
MMLU-Pro	–	–	–	–	86.7%#39	–	–	Vals AI
LegalBench	–	–	–	–	–	85.1%#25	–	Vals AI
CorpFin	–	–	–	–	65.3%#29	–	–	Vals AI
TaxEval	–	–	–	–	76.2%#6	–	–	Vals AI
Excel Modeling Benchmarknot in index	–	–	–	–	–	66.2%#23	–	Vals AI
MedCodenot in index	–	–	–	–	43.4%#45	–	–	Vals AI
MedScribenot in index	–	–	–	–	82.9%#48	–	–	Vals AI
Public Benefits Bench v1.1not in index	–	–	–	–	–	62.4%#27	–	Vals AI
GDP.pdfnot in index	–	–	24.7%#17	–	–	–	24.7%#17	Epoch AI Benchmarking Hub
AA-Omniscience: accuracynot in index	36.8%#146	43.7%#97	44.6%#93	45.5%#89	45.5%#88	46.8%#79	–	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	5.0%#524	10.1%#418	10.2%#412	10.2%#414	11.0%#399	12.1%#382	–	Artificial Analysis

Instruction following

Benchmark	low	medium	high	xhigh	max	Source
LiveBench Languagenot in index	–	–	–	–	82.9%#22	LiveBench
LiveBench Instruction Followingnot in index	–	–	–	–	64.6%#46	LiveBench
IFBench	59.7%#128	62.2%#119	64.4%#108	66.3%#97	71.2%#58	Artificial Analysis
Arena Instruction Followingnot in index	–	–	–	1464#47	–	Arena (LMArena)

Human preference

Benchmark	xhigh	not stated	Source
Arena Text	1467#49	–	Arena (LMArena)
EQ-Bench 4effort not stated	–	1234#11	EQ-Bench
Arena Creative Writingnot in index	1429#73	–	Arena (LMArena)
Arena Multi-turnnot in index	1472#56	–	Arena (LMArena)
Arena Longer Queriesnot in index	1468#62	–	Arena (LMArena)

Multimodal

Benchmark	no reasoning	low	medium	high	xhigh	max	not stated	Source
Arena Vision	–	–	–	–	1271#43	–	–	Arena (LMArena)
LiveBench Designnot in indexeffort not stated	–	–	–	–	–	–	76.0%#26	LiveBench
MMMU-Pro	66.7%#169	76.1%#93	76.8%#86	79.1%#64	79.5%#60	80.7%#44	–	Artificial Analysis
MMMU-Pro (Vals)not in index	–	–	–	–	–	86.5%#20	–	Vals AI
MortgageTaxnot in index	–	–	–	–	67.3%#26	–	–	Vals AI
SAGEnot in index	–	–	–	–	–	47.0%#33	–	Vals AI
Blueprint-Bench 2not in indexeffort not stated	–	–	–	–	–	–	30.8%#17	Epoch AI Benchmarking Hub

Long context

Benchmark	no reasoning	low	medium	high	xhigh	max	Source
AA-LCR	58.7%#287	71.3%#203	74.0%#168	77.7%#128	79.0%#106	83.0%#26	Artificial Analysis
Arena Documentnot in index	–	–	–	–	1477#14	–	Arena (LMArena)
MLCRnot in index	14.4%#84	20.0%#47	22.8%#34	18.9%#55	22.8%#34	27.2%#22	Artificial Analysis

Composite

Benchmark	no reasoning	low	medium	high	xhigh	max	not stated	Source
Epoch Capabilities Indexnot in indexeffort not stated	–	–	–	–	–	–	160.1#13	Epoch AI Benchmarking Hub
LiveBench 2.0effort not stated	–	–	–	–	–	–	75.0%#17	LiveBench
LiveBench 1.0not in index	–	–	–	–	–	77.9%#21	–	LiveBench
AA Intelligence Index v4.3.2	20.8#223	27.5#143	30.1#123	34.2#97	38.0#80	42.1#55	–	Artificial Analysis
Vals Multimodal Indexnot in index	–	–	–	–	65.1%#14	–	–	Vals AI

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
OpenAI	97 tok/s	1.31 s	$4.00	$24.00	1.1M	–
Amazon Bedrock	95 tok/s	1.72 s	$2.20	$13.20	1.1M	–
OpenAI	87 tok/s	2.04 s	$1.00	$6.00	1.1M	–
OpenAI	73 tok/s	1.36 s	$2.00	$12.00	1.1M	–
Azure	61 tok/s	1.90 s	$2.00	$12.00	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	$0.0044	$0.0039	1.6 min
Summarize a 30-page report	12,000 / 600	$0.031	$0.015	1.6 min
Code edit	6,000 / 1,500	$0.030	$0.022	1.7 min
Agentic coding session	60,000 / 4,000	$0.168	$0.087	2.1 min
Structured extraction	2,000 / 200	$0.0064	$0.0037	1.6 min

New this month

Compare GPT-5.6 Terra with…

Questions people ask

What are GPT-5.6 Terra's benchmark scores?: As of 24 Oct 2026, GPT-5.6 Terra (max effort) scores 64.4 ±5.0 on the BenchLeader Index, top 100, strongest in composite (77) and weakest in knowledge (57). That rests on 22 independent benchmark results, and every score on this page links to its publisher.
How does GPT-5.6 Terra rank?: GPT-5.6 Terra (max effort) is #51 of 785 ranked configurations on 24 Oct 2026 (top 100). 142 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
How much does GPT-5.6 Terra cost per 1M tokens?: As of 24 Oct 2026, GPT-5.6 Terra costs $2.00 per 1M input tokens and $12.00 per 1M output tokens, $4.50 blended (three input tokens per output token).
Is GPT-5.6 Terra better than Gemini 3.8 Flash?: As of 24 Oct 2026, Gemini 3.8 Flash (high) and GPT-5.6 Terra (max) are neck and neck on quality (64.4 vs 64.4 on the BenchLeader Index, inside the ±6.3 margin of error), and Gemini 3.8 Flash (high) costs 3.0× less, which makes it the better buy for most workloads. The head-to-head page has the benchmark-by-benchmark detail.
When was GPT-5.6 Terra released?: OpenAI released GPT-5.6 Terra on 9 Jul 2026. BenchLeader has results for it at 7 reasoning-effort settings.

Data as of 24 Oct 2026. Compare these configurations.

Cite as: BenchLeader, “GPT-5.6 Terra: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/gpt-5-6-terra.