Skip to main content
BenchLM
Advertise

Model comparison

Grok 4.3 vs Qwen3.5 397B

Data verified

Head-to-head evidence from 22 shared benchmark results across 6 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

65.1/100
Margin
8.1pts
← winning
57.01/100
0 category wins4 category wins

Public leaderboard positions: Grok 4.3 #31 (Supported); Qwen3.5 397B #71 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Grok 4.3 and Qwen3.5 397B share 22 comparable benchmark results. 4 of 8 categories are comparable. 3 results are unique to Grok 4.3; 33 to Qwen3.5 397B.

Updated July 23, 2026
Shared results
22
Grok 4.3 only
3
Qwen3.5 397B only
33
Comparable categories
4 / 8

Pick Grok 4.3 if you want the stronger benchmark profile. Qwen3.5 397B only becomes the better choice if coding is the priority or you would rather avoid the extra latency and token burn of a reasoning model.

Confidence note. This is a partial-evidence comparison with 22 shared benchmark results across 6 evidence categories; 4 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

Grok 4.3 is clearly ahead on the BenchAlign aggregate, 65.1 to 57.01. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

Qwen3.5 397B is also the more expensive model on tokens at $0.60 input / $3.60 output per 1M tokens, versus $1.25 input / $2.50 output per 1M tokens for Grok 4.3. Grok 4.3 is the reasoning model in the pair, while Qwen3.5 397B is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. Grok 4.3 gives you the larger context window at 1M, compared with 128K for Qwen3.5 397B.

Category breakdown

Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.

Category scores and score margins for Grok 4.3 and Qwen3.5 397B
CategoryGrok 4.3ΔQwen3.5 397B
CodingGrok 4.347.3Margin→ 19.2Qwen3.5 397B66.5
KnowledgeGrok 4.342.4Margin→ 14.2Qwen3.5 397B56.6
Inst. FollowingGrok 4.381.3Margin→ 11.3Qwen3.5 397B92.6
MultimodalGrok 4.378.1Margin→ 1.5Qwen3.5 397B79.6
AgenticGrok 4.3Not measuredMarginNo overlapQwen3.5 397B56.5
ReasoningGrok 4.3Not measuredMarginNo overlapQwen3.5 397B63.2
MathGrok 4.3Not measuredMarginNo overlapQwen3.5 397B90.6
MultilingualGrok 4.3Not measuredMarginNo overlapQwen3.5 397B84.7

Decisive benchmark drivers

The largest measured benchmark gaps in this matchup, with exact reported values.

More
A · Grok 4.3B · Qwen3.5 397B

  1. HLE

    Knowledge
    Source ↗
    A 35%B 28.7%
    Winner: Grok 4.3Δ 6.3
    HLE: Grok 4.3 scored 35%; Qwen3.5 397B scored 28.7%. Grok 4.3 wins this benchmark.

  2. GPQA

    Knowledge
    Source ↗
    A 90.1%B 88.4%
    Winner: Grok 4.3Δ 1.7
    GPQA: Grok 4.3 scored 90.1%; Qwen3.5 397B scored 88.4%. Grok 4.3 wins this benchmark.

  3. MMMU-Pro

    Multimodal
    Source ↗
    A 78.1%B 79%
    Winner: Qwen3.5 397BΔ 0.9
    MMMU-Pro: Grok 4.3 scored 78.1%; Qwen3.5 397B scored 79%. Qwen3.5 397B wins this benchmark.

Operational comparison

Runtime and commercial metrics are compared only when both models have a complete sourced value.

MetricGrok 4.3Qwen3.5 397BComparison
Input / output priceUSD per 1M tokensGrok 4.3$1.25 input / $2.5 outputQwen3.5 397B$0.6 input / $3.6 outputGrok 4.3 has the lower combined listed price.
Generation speedtokens per secondGrok 4.3209 tok/sQwen3.5 397B96 tok/sGrok 4.3 has the higher measured throughput.
First-answer latencyseconds to first tokenGrok 4.312.36 sQwen3.5 397B2.44 sQwen3.5 397B reaches the first token sooner.
Context windowmaximum listed tokensGrok 4.31MQwen3.5 397B128KGrok 4.3 lists the larger context window.

Benchmark Deep Dive

Agentic
BenchmarkGrok 4.3Qwen3.5 397BResult
τ²-bench resultsSource 97.7%95.6%Grok 4.3 leads
GDPval-AASource 29.2%23.1%Grok 4.3 leads
AA Agentic IndexSource 24.1%19.9%Grok 4.3 leads
APEX-Agents-AASource 17.0%15.3%Grok 4.3 leads
GDPval-AASource 1085962Grok 4.3 leads
Gert LabsSource 43.86%46.76%Qwen3.5 397B leads
ResearchClawBenchSource 12.4%14.2%Qwen3.5 397B leads
Terminal-Bench 2.0Source —52.5%Not comparable
BrowseCompSource —62%Not comparable
Claw-EvalSource —56.8%Not comparable
QwenClawBenchSource —51.8%Not comparable
τ³-bench resultsSource —68.4%Not comparable
VITA-BenchSource —43.7%Not comparable
DeepPlanningSource —37.6%Not comparable
ToolathlonSource —36.3%Not comparable
MCP AtlasSource —46.1%Not comparable
MCP-TasksSource —74.2%Not comparable
WideResearchSource —74.0%Not comparable
CodingQwen3.5 397B wins
BenchmarkGrok 4.3Qwen3.5 397BResult
SciCodeSource 47.3%—Not comparable
AA Coding IndexSource 42.3%48.2%Qwen3.5 397B leads
AA-SciCodeSource 47.3%42.0%Grok 4.3 leads
SWE-bench VerifiedSource —76.2%Not comparable
LiveCodeBench v6Source —83.6%Not comparable
SWE-bench ProSource —50.9%Not comparable
Reasoning
BenchmarkGrok 4.3Qwen3.5 397BResult
AA-LCRSource 64.3%65.7%Qwen3.5 397B leads
CritPtSource 8.0%1.7%Grok 4.3 leads
LongBench v2Source —63.2%Not comparable
AI-NeedleSource —68.7%Not comparable
KnowledgeQwen3.5 397B wins
BenchmarkGrok 4.3Qwen3.5 397BResult
Artificial Analysis Intelligence IndexSource 37.6%33.7%Grok 4.3 leads
GPQASource 90.1%88.4%Grok 4.3 leads
HLESource 35%28.7%Grok 4.3 leads
AA-Omniscience AccuracySource 34.6%31.4%Grok 4.3 leads
AA-Omniscience Hallucination RateSource 25.0%89.1%Grok 4.3 leads
AA-GPQA DiamondSource 90.1%89.3%Grok 4.3 leads
AA-HLESource 35.0%27.3%Grok 4.3 leads
AA-Omniscience IndexSource 18.3%-29.8%Grok 4.3 leads
SuperGPQASource —70.4%Not comparable
MMLU-ProSource —87.8%Not comparable
MMLU-ReduxSource —94.9%Not comparable
C-EvalSource —93%Not comparable
Math
BenchmarkGrok 4.3Qwen3.5 397BResult
AIME26Source —93.3%Not comparable
HMMT Feb 2025Source —94.8%Not comparable
HMMT Nov 2025Source —92.7%Not comparable
HMMT Feb 2026Source —87.9%Not comparable
MMAnswerBenchSource —80.9%Not comparable
Multilingual
BenchmarkGrok 4.3Qwen3.5 397BResult
MMLU-ProXSource —84.7%Not comparable
NOVA-63Source —59.1%Not comparable
MultimodalQwen3.5 397B wins
BenchmarkGrok 4.3Qwen3.5 397BResult
MMMU-ProSource 78.1%79%Qwen3.5 397B leads
Design Arena WebsiteSource 1225—Not comparable
AA-MMMU-ProSource 78.1%77.3%Grok 4.3 leads
MathVisionSource —88.6%Not comparable
CharXivSource —80.8%Not comparable
VideoMMMUSource —84.7%Not comparable
ScreenSpot ProSource —65.6%Not comparable
V*Source —95.8%Not comparable
Inst. FollowingQwen3.5 397B wins
BenchmarkGrok 4.3Qwen3.5 397BResult
IFBenchSource 81.3%—Not comparable
AA-IFBenchSource 81.3%78.8%Grok 4.3 leads
IFEvalSource —92.6%Not comparable
Frequently Asked Questions (5)

Which is better, Grok 4.3 or Qwen3.5 397B?

Grok 4.3 is ahead on BenchLM's BenchAlign leaderboard, 65.1 to 57.01. The biggest single separator in this matchup is HLE, where the scores are 35% and 28.7%.

Which is better for knowledge tasks, Grok 4.3 or Qwen3.5 397B?

Qwen3.5 397B has the edge for knowledge tasks in this comparison, averaging 56.6 versus 42.4. Inside this category, AA-Omniscience Hallucination Rate is the benchmark that creates the most daylight between them.

Which is better for coding, Grok 4.3 or Qwen3.5 397B?

Qwen3.5 397B has the edge for coding in this comparison, averaging 66.5 versus 47.3. Inside this category, AA Coding Index is the benchmark that creates the most daylight between them.

Which is better for multimodal and grounded tasks, Grok 4.3 or Qwen3.5 397B?

Qwen3.5 397B has the edge for multimodal and grounded tasks in this comparison, averaging 79.6 versus 78.1. Inside this category, MMMU-Pro is the benchmark that creates the most daylight between them.

Which is better for instruction following, Grok 4.3 or Qwen3.5 397B?

Qwen3.5 397B has the edge for instruction following in this comparison, averaging 92.6 versus 81.3. Inside this category, AA-IFBench is the benchmark that creates the most daylight between them.

Related Comparisons

Explore More

xAI Compare Pricing Methodology Find Your Best LLM Overall Rankings
Last updated: July 23, 2026

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.