VerdictPal · editorial desk · 2026VerdictPal

Compare · Two models, head-to-head

Pick two models. See the diff.

Composite scores, benchmark ranks, capability grids, and pricing per million tokens. All sourced from the same data that powers the model atlas.

Cards Models

Grok 4.3 vs Gemini 3.1 Pro

Model A

Model B

MODEL A

Grok 4.3

xAI · Grok

Composite 60Intel 4717 rows9 sources

xAI's April 2026 frontier model with a 2M-token context window, real-time X search built in, and Vals Index 46.63%. Not the smartest model on the board, but the only one with a 2M context and a live social graph.

MODEL B

Gemini 3.1 Pro

Google · Gemini

Composite 76Intel 5316 rows9 sources

Google's February 2026 frontier model. AA Intelligence Index 57.2 (#4). SWE-bench Verified 80.6%, Terminal-Bench 2.0 78.4%, ARC-AGI-2 77.1% (Google's top non-flash model). 1M context, $7 / $21 per 1M tokens.

Grok 4.34stats won

Tied0stats

Gemini 3.1 Pro3stats won

Spec showdown

Capacity, price, freshness.

Grok 4.3vsGemini 3.1 Pro

Gemini 3.1 Pro leads on composite score (76 vs 60), with deeper benchmark coverage across more families. Gemini 3.1 Pro has the higher Vals/AA intelligence score (53 vs 47).

Composite score
Grok 4.360Gemini 3.1 Pro76Winner on this row
Intelligence (Vals/AA)
Grok 4.347Gemini 3.1 Pro53Winner on this row
Context window
Grok 4.32MWinner on this rowGemini 3.1 Pro1M
Max output
Grok 4.3128KWinner on this rowGemini 3.1 Pro66K
Input price / 1M (lower wins)
Grok 4.3$5.0Winner on this rowGemini 3.1 Pro$7.0
Output speed (AA default API)
Grok 4.3124 t/sWinner on this rowGemini 3.1 Pro—
Time to first token (AA default API)
Grok 4.331.32sGemini 3.1 Pro—Winner on this row

Composite score

Intelligence (Vals/AA)

Context window

Max output

128K

66K

Input price / 1M (lower wins)

$5.0

$7.0

Output speed (AA default API)

124 t/s

—

Time to first token (AA default API)

31.32s

—

Capability grid

What each model can do.

Modalities

Grok 4.3Gemini 3.1 Pro

Text●●

Image●●

Audio—●

Video—●

Code●●

Tool use●●

Reasoning●●

Access

Grok 4.3Gemini 3.1 Pro

API●●

Consumer app●●

Open weights——

Managed cloud●●

On-device——

Family radar

Top score per benchmark family.

Knowledge

Reasoning

Coding

Agentic

Long context

Tool use

Human preference

Benchmark ranks

Every shared benchmark, side by side.

Knowledge

MMLU#11/22 · #5/22

88.5%

90.3%

Reasoning

GPQA Diamond#31/69 · #52/69

88%

78.4%

LiveBench#17/21 · #7/21

60.8%

65.8%

Coding

SWE-bench Verified#19/23 · #8/23

77.6%

80.6%

Agentic

Terminal-Bench#33/69 · #30/69

63.8%

65.4%

Vals Index#15/20 · #11/20

46.63%

53.42%

Long context

LongBench v2#17/19 · #4/19

46.2%

55.4%

Tool use

BFCL#12/20 · #7/20

77.2%

81.6%

Human preference

LMArena (Chatbot Arena)#4/13 · #9/13

1441

1402

Dossier facts

Dates, sources, fineprint.

Grok 4.3Gemini 3.1 Pro

ProviderxAIGoogle

FamilyGrokGemini

Release date2026-04-302026-02-19

Knowledge cutoff2026-03-152026-01-31

Context window2M1M

Max output128K66K

Input price /1M$5.0$7.0

Output price /1M$25$21

Statussolidflagship

Composite score6076

Intelligence score4753

Intel sourcevalsvals

Benchmark rows1716

Source count99

Freshnessfreshfresh

Pricing checked2026-06-012026-06-01

Ecosystem

Which tools wrap each model.

Grok 4.3

Not listed in any tool yet.

Gemini 3.1 Pro

Verdict

Which model to actually pick.

Gemini 3.1 Pro leads on composite score (76 vs 60), with deeper benchmark coverage across more families. Gemini 3.1 Pro has the higher Vals/AA intelligence score (53 vs 47).

Open Grok 4.3's dossier Open Gemini 3.1 Pro's dossier