Skip to main content

Model profile

Grok 4.20

xAICurrentReleased Mar 10, 2026
Data verified
Superseded:xAI has released newer models in this line —Grok 4.3·Grok 4.5

Grok 4.20 by xAI scores 54.68/100 on the public leaderboard (#88 of 200), with an Arena Elo of 1474 and a 2M-token context window. API pricing is $2/$6 per million input/output tokens. Newer replacements: Grok 4.3 and Grok 4.5.

Overall Score
54.68Public #88 of 200
Arena Elo
1474
Eligible category ranks
3of 8
Price (1M tokens)
$2 in / $6 out
API pricing
Speed
233tok/s
Context
2M

Evidence coverage

18 of 323 tracked benchmarks are published. 2 are verified and 16 provisional. 5 of 8 categories are measured.

Updated July 23, 2026Methodology
Published / tracked
18 / 323
Verified
2
Provisional
16
Categories with evidence
5 / 8

Evidence by category

  • Agentic3 benchmarks
    Mixed evidence
  • Coding4 benchmarks
    Mixed evidence
  • Reasoning1 benchmark
    Reported
  • Knowledge4 benchmarks
    Reported
  • Math0 benchmarks
    Not measured
  • Multilingual0 benchmarks
    Not measured
  • Multimodal6 benchmarks
    Reported
  • Inst. Following0 benchmarks
    Not measured
ProprietaryReasoning
Confidence:
Medium
reasoning

Grok 4.20 ranks #88 out of 200 models on the public leaderboard with an overall score of 54.68/100. It does not yet have enough sourced coverage for BenchLM's verified leaderboard. While not a frontier model, it offers specific advantages depending on the use case.

Grok 4.20 is a proprietary model with a 2M token context window. It uses explicit chain-of-thought reasoning, which typically improves performance on math and complex reasoning tasks at the cost of higher latency and token usage.

Grok 4.20 sits inside the Grok 4.20 family alongside Grok 4.20 Multi-agent. BenchLM links it directly to Grok 4.1 as the earlier related model in that lineage. This profile currently has 18 of 323 tracked benchmarks. BenchLM only exposes non-generated benchmark rows publicly, so missing categories stay blank until a sourced evaluation is available.

Its strongest category is Multimodal & Grounded (#25), while its weakest is Coding (#83). This performance profile makes it particularly strong for screenshots, documents, charts, and grounded multimodal workflows.

Peer position

Exact provisional scores and ranks for the closest listed peers. A score can appear before a model clears the evidence threshold for a rank, so equal scores can have different rank states.

Range 54.4855.15

  1. GPT-5 (medium)
    OpenAI
    #8455.15
    GPT-5 (medium) is #84 with a score of 55.15.
    Compare
  2. GLM-4.6
    Z.AI
    #8555.12
    GLM-4.6 is #85 with a score of 55.12.
    Compare
  3. Step 3.5 Flash
    StepFun
    #8655.1
    Step 3.5 Flash is #86 with a score of 55.1.
    Compare
  4. Kimi K2.7 Code
    Moonshot AI
    #8755.0
    Kimi K2.7 Code is #87 with a score of 55.0.
    Compare
  5. Grok 4.20Current model
    xAI
    #8854.68
    Grok 4.20 is #88 with a score of 54.68.
  6. DeepSeek LLM 2.0
    DeepSeek
    #8954.53
    DeepSeek LLM 2.0 is #89 with a score of 54.53.
    Compare
  7. GPT-5.1-Codex-Max
    OpenAI
    #9054.48
    GPT-5.1-Codex-Max is #90 with a score of 54.48.
    Compare

Category percentile

More

Relative position among models eligible for each sourced category. A higher percentile means a stronger position within that category's ranked cohort; 100 is highest.

  1. Multimodal14%
    Eligible cohort rank #25 of 29Category score 34.9
  2. Agentic59%
    Eligible cohort rank #49 of 119Category score 49.6
  3. Coding32%
    Eligible cohort rank #83 of 122Category score 47.0

Category evidence

Scores and ranks appear only where this model has published benchmark evidence. Categories without displayable source records remain not measured.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticRank #49 of 119Percentile 59thWeight 22%3 benchmarksMixed sources49.6
CodingRank #83 of 122Percentile 32ndWeight 20%4 benchmarksMixed sources47.0
ReasoningRank Not rankedWeight 17%1 benchmarkReported65.2
KnowledgeWeight 12%4 benchmarksReportedScore pending
MathWeight 5%0 benchmarksNot measuredNot measured
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalRank #25 of 29Percentile 14thWeight 12%6 benchmarksReported34.9
Inst. FollowingWeight 5%0 benchmarksNot measuredNot measured

Chatbot Arena performance

Scroll horizontally to inspect confidence intervals and vote counts.

Chatbot Arena Elo, confidence interval, and vote count by evaluation view
ViewEloConfidence intervalVotes
Text Overall1474±4.726,827
Coding1508±7.77,213
Math1452±14.81,601
Instruction Following1449±7.28,360
Creative Writing1460±10.03,970
Multi-turn1482±9.64,384
Hard Prompts1486±5.616,642
Hard Prompts (English)1487±7.28,378
Longer Query1463±7.09,785

Benchmark Details

Rows below have a displayable published verification record. Each source link and provenance note remains in the page HTML while its category is closed. Source-unverified manual rows and generated rows stay hidden.

Agentic3 benchmarks
Terminal-Bench 2.0Secondary exact
47.1%Weighted 38%
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
DeepSearchQASecondary exact
62.8%Display only
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
Gert LabsBenchmark exact

Gert Labs Composite Game Benchmark

38.36%Display only
Source: Gert Labs rankingsProvenance: Gert Labs reports this composite leaderboard score in the public rankings API. BenchLM scales the source gscore from 0-1 to 0-100 and stores it as a display-only agentic benchmark.
Coding4 benchmarks
SWE-bench VerifiedSecondary exact

Software Engineering Benchmark Verified

76.7%Weighted 16%
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
SWE-bench ProSecondary exact
51.8%Weighted 10%
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
LiveCodeBench ProSecondary exact
74.2%Display only
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
Vibe Code BenchBenchmark exact

Vibe Code Bench v1.1

4.06%Display only
Source: Vals AI: Vibe Code Bench v1.1Provenance: Vals Vibe Code Bench v1.1 reports this exact row under grok/grok-4.20-0309-reasoning; BenchLM stores it on the local vibeCodeBench key.
Reasoning1 benchmark
ARC-AGI-2Secondary exact

Abstraction and Reasoning Corpus for AGI v2

53.3%Weighted 31%
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
Knowledge4 benchmarks
GPQA-DSecondary exact

GPQA Diamond

88.5%Display only
Source: Meta AI: Muse Spark comparison chartProvenance: BenchLM maps Meta’s “Grok 4.2 Reasoning” chart row onto the tracked Grok 4.20 reasoning family entry.
HLE w/o toolsSecondary exact

Humanity's Last Exam without tools

31.6%Display only
Source: Meta AI: Muse Spark comparison chartProvenance: BenchLM maps Meta’s “Grok 4.2 Reasoning” chart row onto the tracked Grok 4.20 reasoning family entry.
HealthBench HardSecondary exact
20.3%Display only
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
MedXpertQA (Text)Secondary exact

MedXpertQA Text

50.2%Display only
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
Multimodal6 benchmarks
MMMU-ProSecondary exact

Massive Multi-discipline Multimodal Understanding Pro

75.2%Weighted 45%
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
CharXivSecondary exact

CharXiv Reasoning

60.9%Weighted 25%
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
ERQASecondary exact
54.1%Display only
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
SimpleVQASecondary exact
57.4%Display only
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
MedXpertQA (MM)Secondary exact

MedXpertQA Multimodal

65.8%Display only
Source: Meta AI: Muse Spark comparison chartProvenance: Secondary exact
Design Arena WebsiteReported

Design Arena Website Elo

1257Display only
Source: OpenRouter model benchmarksProvenance: Display-only Design Arena Website Elo synced from OpenRouter model benchmark metadata. It is excluded from BenchLM weighted scoring.

Grok 4.20 Family

Reasoning

Related Earlier Model

Grok 4.1

Frequently Asked Questions

How does Grok 4.20 perform overall in AI benchmarks?

Grok 4.20 has 18 published benchmark scores on BenchLM, but it does not yet have enough non-generated coverage to receive a global overall rank.

Is Grok 4.20 good for knowledge and understanding?

Grok 4.20 has visible benchmark coverage in knowledge and understanding, but BenchLM does not currently assign it a global category rank there.

Is Grok 4.20 good for coding and programming?

Grok 4.20 ranks #83 out of 122 models in coding and programming benchmarks with an average score of 47. There are stronger options in this category.

Is Grok 4.20 good for reasoning and logic?

Grok 4.20 has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is Grok 4.20 good for agentic tool use and computer tasks?

Grok 4.20 ranks #49 out of 119 models in agentic tool use and computer tasks benchmarks with an average score of 49.6. There are stronger options in this category.

Is Grok 4.20 good for multimodal and grounded tasks?

Grok 4.20 ranks #25 out of 29 models in multimodal and grounded tasks benchmarks with an average score of 34.9. There are stronger options in this category.

Which sibling models are related to Grok 4.20?

Grok 4.20 belongs to the Grok 4.20 family. Related variants on BenchLM include Grok 4.20 Multi-agent.

Does Grok 4.20 have full benchmark coverage on BenchLM?

Not yet. Grok 4.20 currently has 18 published benchmark scores out of the 323 benchmarks BenchLM tracks. BenchLM only exposes non-generated public benchmark rows, so missing categories stay blank until a sourced evaluation is available.

What is the context window size of Grok 4.20?

Grok 4.20 has a published context window of 2M, which determines how much text it can process in a single interaction.

Last updated: July 23, 2026 · Runtime metrics stay blank until BenchLM has a sourced snapshot.

Choose with this week’s evidence

Join 2,000+ readers for ranking moves, new releases, pricing changes, and the evidence behind them.

Free. One email per week.