ZMIME
Comparison · 3 models · Updated Nov 22, 2026

Gemini 3.1 Pro Preview vs GPT-5.4 vs GPT-6 Astra

Too close to call on our weighted score (Gemini 3.1 Pro Preview 68, GPT-6 Astra 66, GPT-5.4 65). The right pick depends on what you value most.GPT-6 Astra comes out ahead, 89 to 84 and 82 on our weighted score, though Gemini 3.1 Pro Preview is 4.4× cheaper per token.

  1. Google

    Gemini 3.1 Pro Preview

    Released Feb 19, 2026Latest

    68/100
    • ECI156.3
    • Price$2.00 / $12.00
    • Context1.05M
  2. OpenAI

    GPT-5.4

    Released Mar 5, 2026

    65/100
    • ECI157.9
    • Price$2.50 / $15.00
    • Context1.05M
  3. OpenAI

    GPT-6 Astra

    Released Sep 4, 2026

    66/100
    • ECI169.2
    • Price$10.00 / $50.00
    • Context1.05M
  4. Google

    Gemini 3.1 Pro Preview

    Released Feb 19, 2026Latest

    84/100
    • ECI156.3
    • Price$2.00 / $12.00
    • Context1.05M
  5. OpenAI

    GPT-5.4

    Released Mar 5, 2026

    82/100
    • ECI157.9
    • Price$2.50 / $15.00
    • Context1.05M
  6. Our pick

    OpenAI

    GPT-6 Astra

    Released Sep 4, 2026

    89/100
    • ECI169.2
    • Price$10.00 / $50.00
    • Context1.05M
01 — Verdict

Too close to callGPT-6 Astra is our pick

It is close. Our weighted score puts them within 1 points (Gemini 3.1 Pro Preview 68/100, GPT-6 Astra 66/100, GPT-5.4 65/100), so choose by what matters most for your work: GPT-6 Astra for raw capability and Gemini 3.1 Pro Preview on price. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.GPT-6 Astra is the better all-round choice, scoring 89/100 against Gemini 3.1 Pro Preview (84) and GPT-5.4 (82). It leads on capability. Gemini 3.1 Pro Preview wins on inputs & features. The score weighs capability 67%, inputs & features 20%, context window 13%.

  • CapabilityGPT-6 AstraCapabilities Index (ECI): GPT-6 Astra 169.2 · GPT-5.4 157.9 · Gemini 3.1 Pro Preview 156.3
  • Lowest priceGemini 3.1 Pro PreviewGemini 3.1 Pro Preview $4.50 · GPT-5.4 $5.63 · GPT-6 Astra $20.00 per 1M tokens (3:1 blend)
  • Longest contextAbout the sameGPT-5.4 1,050,000 · GPT-6 Astra 1,050,000 · Gemini 3.1 Pro Preview 1,048,576 tokens
  • Widest inputsGemini 3.1 Pro PreviewGemini 3.1 Pro Preview: Text, Images, PDFs, Audio, Video · GPT-5.4: Text, Images, PDFs · GPT-6 Astra: Text, Images, PDFs
  • Self-hostingNo confirmed open weightsAll three are available only through APIs
How the score is built
MeasureWeightGemini 3.1 Pro PreviewGPT-5.4GPT-6 Astra
CapabilityCapabilities Index (ECI)50%848596
Price25%18140
Inputs & features15%1008080
Context window10%656565
Overall100%68/10065/10066/100
How the score is built
MeasureWeightGemini 3.1 Pro PreviewGPT-5.4GPT-6 Astra
CapabilityCapabilities Index (ECI)67%848596
Inputs & features20%1008080
Context window13%656565
Overall100%84/10082/10089/100
02 — Side by side

Every spec in one table

Highlighted cells lead their row. Unavailable specifications are marked “Not published”. Publisher results include their evaluation settings and source.

Gemini 3.1 Pro Preview vs GPT-5.4 vs GPT-6 Astra specifications side by side
SpecificationGemini 3.1 Pro PreviewGoogleGPT-5.4OpenAIGPT-6 AstraOpenAI
Capability
Capabilities Index (ECI)156.3157.9169.2 (best)
ECI rank#41 of 159#29 of 159#4 of 159 (best)
ECI confidence interval154.5–158.2156.0–160.0166.4–172.5
ECI evaluation dateFeb 19, 2026Mar 5, 2026Sep 3, 2026
Documented capabilitiesReasoning-first Gemini preview for agentic coding and complex problem solvingAgent-ready GPT for coding and computer-use workflows at a lower costGPT-6 Astra is OpenAI's most capable model for complex reasoning, coding, computer use, research, and document creation.
Model parametersNot publishedNot publishedNot published
Output modalitiesTextTextText
GPQA DiamondGraduate-level science questions94.4%high93.3%xhigh95.8%max (best)
FrontierMath Tiers 1–3Research-level mathematics59.7%78.6%xhigh93.7%max (best)
OTIS Mock AIME 2024–2025Competition mathematics95.6%97.8%high100%max (best)
SWE-bench VerifiedFixing real GitHub issues75.6%76.9%high (best)Not published
SimpleQA VerifiedShort factual questions73.5%high45.1%xhigh75.6%max (best)
Publisher-reported benchmarks
SWE-Bench ProPublisher-reported result54.2resolve rateReported resultMay 28, 2026Not publishedNot published
Terminal-Bench 2.1Publisher-reported result70.3success rateReported resultMay 28, 2026Not publishedNot published
SWE-Bench Pro · publicPublisher-reported result46.1resolve rateReported result59.1resolve rateReported resultNot published
SWE-Atlas Codebase QnAPublisher-reported result13.5scoreReported result40.8scoreReported resultNot published
SWE-Atlas RefactoringPublisher-reported result33.81scoreReported result44.29scoreReported resultNot published
SWE-Atlas Test WritingPublisher-reported result29.84scoreReported result44.36scoreReported resultNot published
Artificial Analysis Coding Agent Indexhigh43average pass@1Reported resultNot publishedNot published
SWE-Atlas Codebase QnAhigh45.6pass@1Reported resultNot publishedNot published
SWE-Bench Pro · hard-aahigh15.1pass@1Reported resultNot publishedNot published
Terminal-Bench 2.1high68.3pass@1Reported resultNot publishedNot published
GPQA DiamondPublisher-reported result94.3accuracyReported resultApr 23, 202692.8accuracyReported resultApr 23, 202696accuracyReported resultSep 3, 2026
Humanity's Last Exam · full set, text + MMPublisher-reported result44.4accuracyReported resultMay 19, 2026Not publishedNot published
ARC-AGI-2Publisher-reported result77.1accuracyReported resultMay 19, 2026Not publishedNot published
MMMU Prono tools80.5accuracyReported resultMay 19, 202681.2accuracyReported resultApr 23, 2026Not published
MCP AtlasPublisher-reported result78.2success rateReported resultMay 19, 2026Not publishedNot published
OSWorld-VerifiedPublisher-reported result76.2success rateReported resultMay 19, 202675success rateReported resultApr 23, 2026Not published
CharXiv Reasoningno tools83.3accuracyReported resultMay 19, 2026Not publishedNot published
GDPval-AAPublisher-reported result1314EloReported resultMay 19, 2026Not publishedNot published
Artificial Analysis Coding Agent IndexmediumNot published53.6average pass@1Reported resultNot published
SWE-Atlas Codebase QnAmediumNot published72.4pass@1Reported resultNot published
SWE-Bench Pro · hard-aamediumNot published18.4pass@1Reported resultNot published
Terminal-Bench 2.1mediumNot published69.8pass@1Reported resultNot published
Terminal-Bench 2.0Publisher-reported resultNot published75.1success rateReported resultApr 23, 2026Not published
Humanity's Last Examno toolsNot published39.8accuracyReported resultApr 23, 2026Not published
Humanity's Last Examwith toolsNot published52.1accuracyReported resultApr 23, 202657.2accuracyReported resultSep 3, 2026
BrowseCompPublisher-reported resultNot published82.7accuracyReported resultApr 23, 202691.5accuracyReported resultSep 3, 2026
GDPvalPublisher-reported resultNot published83wins or tiesReported resultApr 23, 2026Not published
ARC-AGI-2VerifiedNot published73.3accuracyReported resultApr 23, 2026Not published
FrontierMath · Tier 1-3Publisher-reported resultNot published47.6accuracyReported resultApr 23, 2026Not published
FrontierMath · Tier 4Publisher-reported resultNot published27.1accuracyReported resultApr 23, 2026Not published
Agents' Last ExamPublisher-reported resultNot publishedNot published59.3Reported resultSep 3, 2026
OSWorld 2.0 · V2-Offline (v2026.08.08)Publisher-reported resultNot publishedNot published72.6partial scoreReported resultSep 3, 2026
ScreenSpot-Prono toolsNot publishedNot published92.7accuracyReported resultSep 3, 2026
AutomationBenchPublisher-reported resultNot publishedNot published41.4success rateReported resultSep 3, 2026
BenchCADwith toolsNot publishedNot published95.9geometric overlapReported resultSep 3, 2026
Terminal-Bench 4.0Publisher-reported resultNot publishedNot published57.9success rateReported resultSep 3, 2026
DeepSWE 1.1Publisher-reported resultNot publishedNot published74.1resolve rateReported resultSep 3, 2026
FrontierCode 1.1 · ExtendedPublisher-reported resultNot publishedNot published64.5scoreReported resultSep 3, 2026
Terminal-Bench Science 0.1Publisher-reported resultNot publishedNot published64.6success rateReported resultSep 3, 2026
FrontierMath v2 · Tier 4Publisher-reported resultNot publishedNot published97.6accuracyReported resultSep 3, 2026
ExploitBenchwithout production safeguardsNot publishedNot published100success rateReported resultSep 3, 2026
SRE-Benchsingle attemptNot publishedNot published88success rateReported resultSep 3, 2026
ARC-AGI-3Publisher-reported resultNot publishedNot published99.9RHAEReported resultSep 3, 2026
Artificial Analysis Intelligence Index 4.1.1maxNot publishedNot published61index scoreReported resultSep 3, 2026
Artificial Analysis Coding Agent Index 1.4maxNot publishedNot published67index scoreReported resultSep 3, 2026
Price per million tokens
Input$2.00 (best)$2.50$10.00
Output$12.00 (best)$15.00$50.00
Cached input$0.20 (best)$0.25$1.00
Blended (3:1)$4.50 (best)$5.63$20.00
Long-context rateOver 200K: $4.00 / $18.00Over 272K: $5.00 / $22.50Over 272K: $20.00 / $75.00
Price sourceOfficial Google APIOfficial OpenAI APIOfficial OpenAI API
Limits
Context window1,048,576 tokens1,050,000 tokens (best)1,050,000 tokens (best)
Max inputNot published922,000 tokens922,000 tokens
Max output65,536 tokens128,000 tokens (best)128,000 tokens (best)
Inputs and features
TextYesYesYes
ImagesYesYesYes
PDFsYesYesYes
AudioYesNoNo
VideoYesNoNo
ReasoningYeslow · medium · highYeslow · medium · high · xhighYeslow · medium · high · xhigh · max
Tool callingYesYesYes
Structured outputYesYesYes
File attachmentsYesYesYes
Temperature controlYesYesNo
Reasoning controleffortefforteffort
Availability
WeightsProprietaryProprietaryProprietary
API model IDgemini-3.1-pro-previewgpt-5.4gpt-6-astra
API providers405056 (best)
ReleasedFeb 19, 2026Mar 5, 2026Sep 4, 2026
Knowledge cutoffJan 2025Aug 31, 2025Apr 30, 2026
Publisher documentationPublisher documentationPublisher documentationPublisher documentation
Source metadata updatedFeb 19, 2026Mar 5, 2026Sep 4, 2026
Data sourcesmodels.dev · Epoch AImodels.dev · Epoch AImodels.dev · Epoch AI
Catalog checkedNov 22, 2026Nov 22, 2026Nov 22, 2026
03 — Cost

What would a month cost?

Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.

  • Gemini 3.1 Pro Preview$44.00
  • GPT-5.4$55.00
  • GPT-6 Astra$200.00
04 — Questions

Which should you choose?

Which is better: Gemini 3.1 Pro Preview, GPT-5.4 or GPT-6 Astra?

It is close. Our weighted score puts them within 1 points (Gemini 3.1 Pro Preview 68/100, GPT-6 Astra 66/100, GPT-5.4 65/100), so choose by what matters most for your work: GPT-6 Astra for raw capability and Gemini 3.1 Pro Preview on price. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.GPT-6 Astra is the better all-round choice, scoring 89/100 against Gemini 3.1 Pro Preview (84) and GPT-5.4 (82). It leads on capability. Gemini 3.1 Pro Preview wins on inputs & features. The score weighs capability 67%, inputs & features 20%, context window 13%.

Which is cheaper, Gemini 3.1 Pro Preview, GPT-5.4 or GPT-6 Astra?

Gemini 3.1 Pro Preview is cheaper at $2.00 input / $12.00 output per million tokens (official Google API price). GPT-5.4 costs $2.50 input / $15.00 output per million tokens (official OpenAI API price); GPT-6 Astra costs $10.00 input / $50.00 output per million tokens (official OpenAI API price). At a typical mix of three input tokens to one output token, that is $4.50 per million tokens for Gemini 3.1 Pro Preview versus $5.63 for GPT-5.4 (1.3× as much) and $20.00 for GPT-6 Astra (4.4× as much).

Which scores higher on benchmarks?

GPT-6 Astra scores higher on the Capabilities Index (ECI): GPT-6 Astra 169.2 (#4 of 159), GPT-5.4 157.9 (#29 of 159) and Gemini 3.1 Pro Preview 156.3 (#41 of 159). Their confidence ranges do not overlap (166.4–172.5 vs 156.0–160.0), so the gap is a real one. On individual benchmarks: GPQA Diamond — GPT-6 Astra 95.8%, Gemini 3.1 Pro Preview 94.4%, GPT-5.4 93.3%; FrontierMath Tiers 1–3 — GPT-6 Astra 93.7%, GPT-5.4 78.6%, Gemini 3.1 Pro Preview 59.7%; OTIS Mock AIME 2024–2025 — GPT-6 Astra 100%, GPT-5.4 97.8%, Gemini 3.1 Pro Preview 95.6%; SimpleQA Verified — GPT-6 Astra 75.6%, Gemini 3.1 Pro Preview 73.5%, GPT-5.4 45.1%.

Which is better for coding?

There are no published SWE-bench Verified results for GPT-6 Astra yet, so there is no like-for-like coding score. On overall capability, GPT-6 Astra leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.

Which has the bigger context window?

Their context windows are effectively the same size: Gemini 3.1 Pro Preview 1,048,576, GPT-5.4 1,050,000 and GPT-6 Astra 1,050,000 tokens. Maximum output per response: Gemini 3.1 Pro Preview up to 65,536, GPT-5.4 up to 128,000, GPT-6 Astra up to 128,000 tokens.

Which can read images, PDFs, audio or video?

Gemini 3.1 Pro Preview accepts text, images, PDFs, audio and video; GPT-5.4 accepts text, images and PDFs; GPT-6 Astra accepts text, images and PDFs. Gemini 3.1 Pro Preview handles the widest range of inputs.

Are any of these open source?

No. Gemini 3.1 Pro Preview, GPT-5.4 and GPT-6 Astra are proprietary and only available through APIs and apps.

Which is newer?

GPT-6 Astra is the newest, released Sep 4, 2026. GPT-5.4 came out Mar 5, 2026; Gemini 3.1 Pro Preview came out Feb 19, 2026. Knowledge cutoff: Gemini 3.1 Pro Preview Jan 2025, GPT-5.4 Aug 31, 2025, GPT-6 Astra Apr 30, 2026.

How do you decide the winner?

This rating uses capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 22, 2026.This rating uses capability (67%, independent benchmark results); inputs & features (20%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (13%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 22, 2026.