BenchLeader

Gemini 4 Carbon vs GPT-6.2 Astra

Full profiles: Gemini 4 Carbon (Google) and GPT-6.2 Astra (OpenAI): every benchmark, price and rank.

Updated · Data from 13 sources

As of 19 Nov 2026, Gemini 4 Carbon and GPT-6.2 Astra are neck and neck on quality (68.0 vs 68.9 on the BenchLeader Index, inside the ±2.9 margin), and Gemini 4 Carbon is 4.0× cheaper at $4.00 vs $16.00 per 1M tokens blended. Choose Gemini 4 Carbon for agents & tools (69 vs 66) and fast first answers (72.4 s vs 202.7 s); choose GPT-6.2 Astra for coding (76 vs 73), reasoning (73 vs 72), and maths (73 vs 71).

Key facts (as of 19 Nov 2026)

  • Gemini 4 Carbon scores 68.0 on the BenchLeader Index and GPT-6.2 Astra scores 68.9 (BenchLeader, 19 Nov 2026).
  • Gemini 4 Carbon costs $4.00 per 1M tokens blended (list price, 19 Nov 2026).
  • GPT-6.2 Astra costs $16.00 per 1M tokens blended (list price, 19 Nov 2026).
  • Of the 51 benchmarks both models have, GPT-6.2 Astra leads 30 (BenchLeader, 19 Nov 2026).
  • On Arena WebDev, Gemini 4 Carbon is rated 1769 and GPT-6.2 Astra 1781 (LMArena, as published by 19 Nov 2026).

Which to pick for…

The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.

Use casePickWhyConfidence
CodingGPT-6.2 Astra76 vs 73 category score; Arena WebDev 1781 vs 1769Medium
Agents & tool useGemini 4 Carbon69 vs 66 category score; Terminal-Bench 64.5% vs 62.1%Medium
MathsGPT-6.2 Astra73 vs 71 category score; FrontierMath Tiers 1–3 92.6% vs 90.5%Medium
ReasoningGPT-6.2 Astra73 vs 72 category score; GPQA Diamond 96.8% vs 96.5%Medium
KnowledgeEither75 vs 74 category score; SimpleQA Verified 77.7% vs 78.5%Either
Chat quality (human preference)–Only Gemini 4 Carbon has an Arena rating (73 category score)Not comparable
Lowest costGemini 4 Carbon$8.00 vs $16.00 per 1M blended; agentic coding session $0.320 vs $0.640–
Fastest first answerGemini 4 Carbon72.4 s (Gemini 4 Carbon (high)) vs 202.7 s (GPT-6.2 Astra (max)) to the first answer token–
High volume on a budgetNeither flagshipA smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs–

Key differences

  • Gemini 4 Carbon (high) and GPT-6.2 Astra (max) are neck and neck on quality (68.0 vs 68.9, within the ±2.9 margin).
  • Gemini 4 Carbon (high) is stronger in agents & tools, knowledge.
  • GPT-6.2 Astra (max) is stronger in coding, composite, maths, reasoning.
  • Gemini 4 Carbon (high) is 4.0× cheaper ($4.00 vs $16.00 per 1M blended).

At a glance

SpecGemini 4 CarbonGPT-6.2 Astra
DeveloperGoogleOpenAI
Released28 Oct 202617 Nov 2026
Context window (tokens)1,048,5761,050,000
Compared atGemini 4 Carbon (high)GPT-6.2 Astra (max)
Open weightsNoNo
Price in / out / cached input per 1M$4.00 / $20.00 / $0.100$8.00 / $40.00 / $0.400

Category by category

Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.

CategoryGemini 4 Carbon (high)GPT-6.2 Astra (max)Leader
Reasoning71.873.1GPT-6.2 Astra
Coding72.876.1GPT-6.2 Astra
Agents & tools69.166.4Gemini 4 Carbon
Maths71.372.5GPT-6.2 Astra
Knowledge74.774.2level
Human preference72.6–not comparable
Multimodal–71.7not comparable
Long context64.3–not comparable
Composite67.869.5GPT-6.2 Astra

On one chart

Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.

2550 avg75ReasoningCodingAgents & toolsMathsKnowledgeGemini 4 Carbon, Reasoning: 71.8Gemini 4 Carbon, Coding: 72.8Gemini 4 Carbon, Agents & tools: 69.1Gemini 4 Carbon, Maths: 71.3Gemini 4 Carbon, Knowledge: 74.7GPT-6.2 Astra, Reasoning: 73.1GPT-6.2 Astra, Coding: 76.1GPT-6.2 Astra, Agents & tools: 66.4GPT-6.2 Astra, Maths: 72.5GPT-6.2 Astra, Knowledge: 74.2

Not drawn, measured for fewer than two of these models: Instruction following, Long context, Multimodal, Human preference.

Model, click to add or removeIndex

Where sources disagree

Across the 51 benchmarks both have, Gemini 4 Carbon leads 18, GPT-6.2 Astra leads 30 and 3 are tied. Independent sources split: Gemini 4 Carbon leads on Vals Index; GPT-6.2 Astra leads on AA Intelligence Index v5.0.1.

Overall indexGemini 4 Carbon (high)GPT-6.2 Astra (max)Leader
BenchLeader Index68.068.9neck and neck
AA Intelligence Index v5.0.160.062.1GPT-6.2 Astra
Vals Index70.567.7Gemini 4 Carbon

Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).

Price, speed and latency

List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.

MeasureGemini 4 Carbon (high)GPT-6.2 Astra (max)
Blended price per 1M tokens$8.00$16.00
Input / output per 1M$4.00 / $20.00$8.00 / $40.00
Cached input per 1M$0.100$0.400
Price per task: chat reply$0.0076$0.015
Price per task: summarize a 30-page report$0.060$0.120
Price per task: code edit$0.054$0.108
Price per task: agentic coding session$0.320$0.640
Price per task: structured extraction$0.012$0.024
Agentic coding session, input mostly cached$0.144$0.298
Cost per benchmark task (AA Index)$7.76$4.00
Output tokens per task (AA Index)123k41k
Of which reasoning72k26k
Time per task (AA Index)38 min14 min
Output speed55 tok/s49 tok/s
Time to first answer72.4 s202.7 s
Time to a complete answer, last measured 13 Nov 2026160.6 s–

Per token is not per task: an average task of Artificial Analysis' Intelligence Index cost $7.76 on Gemini 4 Carbon (high) and $4.00 on GPT-6.2 Astra (max), and Gemini 4 Carbon (high) wrote 3.0× as many output tokens for the same work, which a per-token price hides.

Each model is compared at its best-ranked reasoning-effort setting. Gemini 4 Carbon (high) scores 68.0 and takes 72 s to its first answer token; Gemini 4 Carbon (medium) scores 64.7 in 33 s, Gemini 4 Carbon (low) scores 59.0 in 9 s. GPT-6.2 Astra (max) scores 68.9 and takes 203 s to its first answer token; GPT-6.2 Astra (xhigh) scores 67.1 in 67 s, GPT-6.2 Astra (medium) scores 65.6 in 4 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.

Every benchmark

111 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.

MetricPublisherGemini 4 Carbon (high)GPT-6.2 Astra (max)
BenchLeader IndexBenchLeader68.068.9
Agents & tools scoreBenchLeader69.166.4
Coding scoreBenchLeader72.876.1
Composite scoreBenchLeader67.869.5
Human preference scoreBenchLeader72.6–
Knowledge scoreBenchLeader74.774.2
Long context scoreBenchLeader64.3–
Maths scoreBenchLeader71.372.5
Reasoning scoreBenchLeader71.873.1
Multimodal scoreBenchLeader–71.7
Blended price $/MBenchLeader$8.00$16.00
Output speedBenchLeader55 tok/s49 tok/s
Time to first answerBenchLeader72.4 s202.7 s
Context windowBenchLeader1,048,576 tokens1,050,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub96.5%96.8%
SimpleBenchSimpleBench92.6%†–
Arena Hard PromptsArena (LMArena)1546–
LiveBench 2.0 ReasoningLiveBench90.6%91.6%
HLE-Diamond (AA)Artificial Analysis69.1%72.2%
ARC-AGI-1ARC Prize97.5%98.5%
ARC-AGI-2ARC Prize92.1%95.8%
ARC-AGI-3ARC Prize96.9%99.8%
CritPt v2 hard subset (AA)Artificial Analysis69.0%71.5%
MysteryMechanismVals AI–53.1%
HLE-Diamond (Scale)Scale AI SEAL–66.9%†
Arena ExpertArena (LMArena)1566–
Chess PuzzlesEpoch AI Benchmarking Hub78.0%78.0%
Mystery Game PuzzlesEpoch AI Benchmarking Hub73.0%95.0%
Coding
SciCodeSciCode64.2%–
Arena CodingArena (LMArena)1567–
Arena WebDevArena (LMArena)17691781
LiveBench 2.0 CodingLiveBench91.0%91.9%
Vibe Code Bench v1.1Vals AI89.1%89.9%
Vibe Code Bench 1-100 v1.1Vals AI49.7%55.2%
SWE Atlas v1.1Scale AI SEAL71.5%†60.5%†
SWE Atlas v1.1: Codebase QnAScale AI SEAL73.4%†67.7%†
SWE Atlas v1.1: Test WritingScale AI SEAL76.7%†62.2%†
SWE Atlas v1.1: RefactoringScale AI SEAL64.3%†51.4%†
MirrorCodeEpoch AI Benchmarking Hub–67.8%
Show all 111 benchmarks
MetricPublisherGemini 4 Carbon (high)GPT-6.2 Astra (max)
Coding
CursorBenchCursor60.4%–
ALE-BenchEpoch AI Benchmarking Hub2374.7–
FrontierSWEEpoch AI Benchmarking Hub61.2%–
Terminal-Bench 4.0 (AA)Artificial Analysis65.2%64.1%
Design Arena WebsiteDesign Arena1304†1339†
FrontierCode 2.0 (AA)Artificial Analysis40.6%45.4%
Agents & tools
Terminal-BenchTerminal-Bench64.5%62.1%
APEX-AgentsMercor84.2%†–
Arena AgentArena (LMArena)16.115.2
LiveBench Long AgenticLiveBench71.8%70.8%
LiveBench BrowserLiveBench80.0%93.0%
HiL-BenchScale AI SEAL50.3%†–
GDPval-AA v2.1Artificial Analysis1733–
Analyst Agent (AA)Artificial Analysis–62.5%
Finance Agent v2.1Vals AI67.2%60.8%
Harvey's Legal Agent Benchmark v1.1Vals AI17.5%16.3%
Legal Research BenchVals AI55.8%42.8%
Robo-benchVals AI–32.7%
Tax Agent BenchVals AI51.8%21.8%
Terminal-Bench 4.0 (Vals)Vals AI63.1%63.1%
Terminal-Bench ScienceVals AI–71.4%
Arena Agent Task CompletionArena (LMArena)17.212.4
PostTrainBenchEpoch AI Benchmarking Hub41.0%–
Vending-Bench 2Andon Labs16484.2†–
SpireClimbEpoch AI Benchmarking Hub–41.3%
AutomationBenchArtificial Analysis77.8%77.3%
GDP.pdf (AA)Artificial Analysis28.4%29.8%
Harvey LAB-AA v1.1Harvey–15.0%
EnterpriseOps-GymArtificial Analysis–29.9%
AA-Briefcase v1.1Artificial Analysis16111643
GDPxGDPx by SemiAnalysis10.9%–
Design Arena AgenticDesign Arena1304†1304
Terminal-Bench-Science 0.1 (AA)Artificial Analysis64.8%78.1%
CWE-Bench (AA)Artificial Analysis–65.0%
DeepSec-Bench (AA)Artificial Analysis–46.8%
CyberGym end-to-end (AA)Artificial Analysis–0.0%
MKA (AA)Artificial Analysis54.3%52.7%
FACTS SearchGoogle94.4%†–
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub90.5%92.6%
FrontierMath Tier 4Epoch AI Benchmarking Hub95.1%100.0%
OTIS Mock AIMEEpoch AI Benchmarking Hub100.0%100.0%
ArXivMathMathArena94.3%–
BrokenArXivMathArena84.4%–
Autoform-HardVals AI71.4%87.5%
FrontierMath ErdősEpoch AI Benchmarking Hub2.9%10.3%
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub77.7%78.5%
LiveBench 2.0 Data AnalysisLiveBench86.5%84.1%
LiveBench ResearchLiveBench92.5%91.5%
AA-Omniscience v1.1Artificial Analysis53.550
PRBench LegalScale AI SEAL53.4%†43.1%†
BioMysteryBenchVals AI–80.7%
Excel Modeling Benchmark v1.1Vals AI76.5%76.0%
DrugDiscoveryBenchScale AI SEAL71.5%65.5%†
EBR-benchEpoch AI Benchmarking Hub87.6%92.4%
GDP.pdfEpoch AI Benchmarking Hub56.4%–
AA-Omniscience v1.1: accuracyArtificial Analysis58.3%68.6%
AA-Omniscience v1.1: non-hallucinationArtificial Analysis88.6%40.9%
FACTS ParametricGoogle81.6%†–
Instruction following
MultiChallengeScale AI SEAL–73.2%†
Arena Instruction FollowingArena (LMArena)1524–
Human preference
Arena TextArena (LMArena)1526–
Arena Creative WritingArena (LMArena)1515–
Arena Multi-turnArena (LMArena)1563–
Arena Longer QueriesArena (LMArena)1537–
Multimodal
LiveBench DesignLiveBench80.2%82.8%
MMMU-ProArtificial Analysis89.3%†89.5%
Humanity's Sixth SenseScale AI SEAL71.1%63.2%†
Blueprint-Bench 2Epoch AI Benchmarking Hub58.6%†–
Furniture AssemblyEpoch AI Benchmarking Hub77.5%89.2%
VIALS-AAHandshake AI64.0%65.8%
AA-3DArtificial Analysis–1318
FACTS MultimodalGoogle60.2%†–
Long context
AA-LCR v1.1Artificial Analysis79.3%–
MLCRArtificial Analysis19.4%39.4%
FACTS GroundingGoogle77.2%†–
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub169.8†174.2†
LiveBench 2.0LiveBench83.3%84.9%
AA Intelligence Index v5.0.1Artificial Analysis60.062.1
AA Progress Index v1Artificial Analysis350.8–
Vals IndexVals AI70.567.7
Vals Finance IndexVals AI67.0%57.3%
Vals RSI IndexVals AI27.7%29.0%
Vals Science IndexVals AI–68.4%
Vals Legal IndexVals AI75.9%69.7%
AA Cyber IndexArtificial Analysis–37.3
Safety & honesty
FORTRESSScale AI SEAL16.4%†–

Data as of 2026-11-19. Best configuration of each model; every score links to its source on the model pages.

What changed

  • GPT-6.2 Astra (max): BenchLeader Index 67.8 → 68.9.
  • Gemini 4 Carbon (high): 2 score revisions; input price moved from $2.00/M to $4.00/M (100% up); output price moved from $10.00/M to $20.00/M (100% up); cost per task moved from $4.90/M to $7.76/M (58% up).
  • GPT-6.2 Astra (max): 6 new results (VIALS-AA, Arena WebDev, Arena Agent…).
  • GPT-6.2 Astra (max): BenchLeader Index 67.2 → 67.8.
  • Gemini 4 Carbon (high): cost per task moved from $5.48/M to $4.90/M (11% down).
  • GPT-6.2 Astra (max): 2 new results (MKA (AA), Terminal-Bench).
  • Gemini 4 Carbon (high): BenchLeader Index 68.9 → 68.3.
  • Gemini 4 Carbon (high): 3 new results (PostTrainBench, ALE-Bench, GDP.pdf).
  • Gemini 4 Carbon (high): 2 score revisions.
  • Gemini 4 Carbon (high): 8 new results (LiveBench 2.0, LiveBench Long Agentic, LiveBench 2.0 Reasoning…).

Rankings: Best LLM overall · Best LLM for coding · Best LLM for agents · Best cheap LLMs

Gemini 4 Carbon vs GPT-6.2 Astra: questions

Is Gemini 4 Carbon better than GPT-6.2 Astra?
Gemini 4 Carbon (high) and GPT-6.2 Astra (max) are neck and neck on quality (68.0 vs 68.9, within the ±2.9 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 19 Nov 2026, so check the category scores for your use.
Gemini 4 Carbon vs GPT-6.2 Astra: which is cheaper, and will that change?
Gemini 4 Carbon is cheaper today: $4.00 against $16.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is Gemini 4 Carbon or GPT-6.2 Astra better for coding?
GPT-6.2 Astra scores higher in coding (76 vs 73 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1769 for Gemini 4 Carbon against 1781 for GPT-6.2 Astra.
Which is faster, Gemini 4 Carbon or GPT-6.2 Astra?
Gemini 4 Carbon streams faster (55 vs 49 output tokens per second, Artificial Analysis 7-day median).
Is Gemini 4 Carbon or GPT-6.2 Astra better for agents and tool use?
Gemini 4 Carbon scores higher in agentic tasks (69 vs 66 on the category index, where 50 is average). On Terminal-Bench, the top shared agents & tools benchmark, it is 64.5% for Gemini 4 Carbon against 62.1% for GPT-6.2 Astra.
More questions (4)
Which is better for long documents?
Both take about 1.1M tokens: Gemini 4 Carbon 1,048,576, GPT-6.2 Astra 1,050,000.
Which has lower latency?
Gemini 4 Carbon answers first: 72.4 s against 202.7 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
Do independent benchmarks agree on Gemini 4 Carbon vs GPT-6.2 Astra?
Across the 51 benchmarks both have, Gemini 4 Carbon leads 18, GPT-6.2 Astra leads 30 and 3 are tied. Independent sources split: Gemini 4 Carbon leads on Vals Index; GPT-6.2 Astra leads on AA Intelligence Index v5.0.1.
When was GPT-6.2 Astra released?
GPT-6.2 Astra was released on 17 Nov 2026.