BenchLeader

Grok 4.20

Reasoning effort
Quality#61 of 333
58.4Index
Top quarter
+8.4 on the median
Price#160 of 249
$1.56per 1M tokens
Dearer than most
1.8× the median
Speed#114 of 215
74tokens/s
Below the median
1.0× the median

10 cheaper ranked models score higher: Gemini 3.7 Flash (62.2, $1.50), Gemini 3.8 Flash (62.0, $1.50), Gemini 3.6 Flash (61.3, $1.50) and 7 more.

As of 8 Nov 2026, Grok 4.20 by SpaceXAI scores 58.4 on the BenchLeader Index (593 ranked configurations), with its best results in instruction following (79) and weakest in agents & tools (50). API pricing: $1.25 per 1M input tokens and $2.50 per 1M output tokens ($1.56 blended at 3:1), list price on 8 Nov 2026. It streams about 74 tokens/s (Artificial Analysis). Released 17 Feb 2026.

Key facts (as of 8 Nov 2026)

  • Grok 4.20 API pricing: $1.25 per 1M input tokens and $2.50 per 1M output tokens ($1.56 blended at 3:1) (list price, 8 Nov 2026).
  • Grok 4.20 streams about 74 tokens/s (Artificial Analysis, 8 Nov 2026).
  • Grok 4.20 was released on 17 Feb 2026 (SpaceXAI).

In detail. Grok 4.20 is a SpaceXAI proprietary reasoning model, released 17 Feb 2026. Its best configuration (thinking reasoning effort) ranks #124 of 593 on the BenchLeader Index at 58.4 ±4.1, in the upper half. The ± is the point: 230 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in instruction following (79) and lowest in agents & tools (50). At $1.56 per million tokens blended it is pricier than most ranked models. Output speed of 74 tokens per second puts it slower than most, with a first answer in 18.7 s. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 6 Nov 2026.

Blended price
$1.56/M
$1.25 in · $2.50 out
via Artificial Analysis
Output speed
74 tok/s
measured by Artificial Analysis
First answer
19 s
first token 0.62 s
Context
2M
via OpenRouter
Full answer
62 s
median, reasoning included
Cost per task
–
Tokens per task
–
Released
17 Feb 2026
How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.
  1. Overall index58
  2. Reasoning61
  3. Coding54
  4. Agents & tools50
  5. Maths54
  6. Knowledge53
  7. Instruction following79
  8. Human preference66
  9. Multimodal58
  10. Long context59

Get an email when SpaceXAI ships a new model

And once when it gets ranked, plus the weekly review on Mondays.

Choose providers and emails

Versions

SpaceXAI has shipped 11 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

ModelReleasedIndexRank
Grok 4.7xhigh21 Sep 202658.3#129
Grok 4.6high11 Aug 202660.6#73
Grok 4.58 Jul 202659.9#90
Grok 4.3high17 Apr 202656.6#174
Grok 4.2high10 Mar 2026––
Grok 4.20thinkingthis page17 Feb 202658.4#124
Grok 4.117 Nov 202557.1#162
Grok 49 Jul 202558.4#123
Grok 317 Feb 202552.4#272
Grok 212 Dec 202442.3#484
Grok 4.20 1–––

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

EffortIndexRankSpeedFirst answerChat reply costCategories
no reasoning––70 tok/s1.24 s$0.0013Agents & tools 48 · Instruction following 51 · Long context 35 · Multimodal 47
thinkingbest58.4#12474 tok/s19 s$0.0013Agents & tools 50 · Coding 54 · Human preference 66 · Instruction following 79 · Knowledge 53 · Long context 59 · Maths 54 · Multimodal 58 · Reasoning 61
not stated––84 tok/s0.62 s$0.0013Agents & tools 51 · Coding 52 · Reasoning 42

How its index has moved

9 Sep 2026 to 4 Nov 2026
505764

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmarkno reasoningthinkingnot statedSourceTrend
GPQA Diamond–89.3%#59–Epoch AI Benchmarking Hub
Arena Hard Prompts–1489#53–Arena (LMArena)
GPQA Diamond (AA)not in index77.6%#22391.1%#50–Artificial Analysis
Humanity's Last Exam (AA)not in index27.9%#17734.5%#132–Artificial Analysis
GPQA Diamond (Vals)not in index–88.6%#33–Vals AI
Kagi LLM Benchmark–75.0%#1244.8%#96Kagi LLM Benchmark
ARC-AGI-1–89.5%#84–ARC Prize
ARC-AGI-2–65.1%#75–ARC Prize
ARC-AGI-3–0.1%#66–ARC Prize
CritPtnot in index0.3%#2836.6%#152–Artificial Analysis
Chess Puzzlesnot in index–29.0%#71–Epoch AI Benchmarking Hub
CL-benchnot in indexeffort not stated––22.2%#3Epoch AI Benchmarking Hub
CL-bench Lifenot in indexeffort not stated––11.9%#12Epoch AI Benchmarking Hub
LMCAnot in indexeffort not stated––38.7%#112Epoch AI Benchmarking Hub
DTBenchnot in indexeffort not stated––90.1%#82Epoch AI Benchmarking Hub
ForecastBenchnot in index–64.5%#2563.3%#44ForecastBench

Coding

Benchmarkthinkingnot statedSourceTrend
WeirdMLeffort not stated–52.3%#66WeirdML
Arena Coding1510#59–Arena (LMArena)
Arena WebDev1375#109–Arena (LMArena)
LiveCodeBench84.3%#39–Vals AI
SWE-bench (Vals)not in index72.2%#55–Vals AI
Code Migrationnot in index0.3%#77–Vals AI
Vibe Code Bench v1.1not in index4.1%#110–Vals AI
ALE-Benchnot in indexeffort not stated–1150.3#53Epoch AI Benchmarking Hub
Design Arena Websitenot in index1227#711218#78Design Arena

Agents & tools

Benchmarkno reasoningthinkingnot statedSourceTrend
Terminal-Bencheffort not stated––57.3%#24Terminal-Bench
Terminal-Bench Hard16.7%#18837.9%#62–Artificial Analysis
τ²-Bench Telecom (AA)not in index59.9%#18393.0%#46–Artificial Analysis
Terminal-Bench 2.1 (Vals)–44.2%#65–Vals AI
HiL-Bencheffort not stated––20.0%#28Scale AI SEAL
APEX-Agents (AA)not in index–14.2%#26–Artificial Analysis
Finance Agent v2not in index–28.5%#74–Vals AI
Harvey's Legal Agent Benchmark v1not in index–0.0%#64–Vals AI
Legal Research Benchnot in index–13.9%#75–Vals AI
Tax Agent Benchnot in index–7.0%#72–Vals AI
Terminal-Bench 2.0 (Vals)not in index–40.5%#32–Vals AI
Vending-Bench 2not in indexeffort not stated––4662.9#38Andon Labs
Design Arena Agenticnot in index–1089#55–Design Arena

Knowledge

BenchmarkthinkingSourceTrend
SimpleQA Verified30.2%#74Epoch AI Benchmarking Hub
MMLU-Pro86.3%#43Vals AI
LegalBench77.7%#106Vals AI
CorpFin63.7%#43Vals AI
TaxEval74.1%#41Vals AI
MedQA94.5%#16Vals AI
CaseLaw v2not in index54.5%#38Vals AI
Excel Modeling Benchmarknot in index11.7%#73Vals AI
MedCodenot in index32.2%#90Vals AI
MedScribenot in index63.4%#102Vals AI
Global-MMLU-Lite (AA)not in index89.5%#14Artificial Analysis

Instruction following

Benchmarkno reasoningthinkingSourceTrend
IFBench49.3%#17981.2%#5Artificial Analysis
Arena Instruction Followingnot in index–1447#69Arena (LMArena)

Human preference

BenchmarkthinkingSourceTrend
Arena Text1472#47Arena (LMArena)
Arena Creative Writingnot in index1445#51Arena (LMArena)
Arena Multi-turnnot in index1479#45Arena (LMArena)
Arena Longer Queriesnot in index1466#67Arena (LMArena)

Multimodal

Benchmarkno reasoningthinkingnot statedSourceTrend
Arena Vision–1262#54–Arena (LMArena)
MMMU-Pro64.9%#19174.6%#120–Artificial Analysis
MMMU-Pro (Vals)not in index–83.5%#32–Vals AI
MortgageTaxnot in index–45.4%#84–Vals AI
Blueprint-Bench 2not in indexeffort not stated––0.0%#31Epoch AI Benchmarking Hub

Long context

Benchmarkno reasoningthinkingSourceTrend
AA-LCR v1.123.3%#43569.0%#241Artificial Analysis
Arena Documentnot in index–1440#37Arena (LMArena)

Composite

Benchmarkno reasoningthinkingnot statedSourceTrend
Epoch Capabilities Indexnot in indexeffort not stated––152.3#54Epoch AI Benchmarking Hub
AA Intelligence Index v4.3.2not in index14.2#30725.7#167–Artificial Analysis
AA Progress Index v1not in index224.6#338281.9#169–Artificial Analysis
Vals Multimodal Indexnot in index–39.1%#32–Vals AI

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

API pricing: where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

ProviderSpeedFirst tokenInput $/MOutput $/MContextQuantisation
SpaceXAI88 tok/s1.27 s$2.50$5.002M–
SpaceXAI67 tok/s0.63 s$1.25$2.502M–
SpaceXAI66 tok/s0.60 s$1.25$2.502M–

Price history

The price above, per 1M tokens, every day it changed.

$0.00$1.65$3.30$4.95$6.60Sep 26Oct 26Nov 26
input outputnow $1.25 in · $2.50 out

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.200 per 1M. Reasoning tokens are not modelled.

WorkloadTokens in / outCostWith cachingTime
Chat reply400 / 300$0.0013$0.000922.8 s
Summarize a 30-page report12,000 / 600$0.017$0.007126.9 s
Code edit6,000 / 1,500$0.011$0.006539.1 s
Agentic coding session60,000 / 4,000$0.085$0.0381.2 min
Structured extraction2,000 / 200$0.0030$0.001421.5 s

More from SpaceXAI

New this month

Compare Grok 4.20 with…

See also

Embed or cite

Show where Grok 4.20 ranks on a model card or README. The badge redraws itself with every daily update, carries the data date and links back to this page. To see ranks like this on OpenRouter and Hugging Face model pages, get the Chrome extension.

[![Grok 4.20 on BenchLeader: #61 overall](https://www.benchleader.com/badge/grok-4-20.svg)](https://www.benchleader.com/models/grok-4-20)

Cite this

The ranking changes daily, so a reference carries the data date. APA style:

BenchLeader. (2026, November 8). Grok 4.20 on the BenchLeader Index (method v1.4) [Data set]. Retrieved November 8, 2026, from https://www.benchleader.com/models/grok-4-20

BibTeX:

@misc{benchleader_grok_4_20_20261108,
  author       = {{BenchLeader}},
  title        = {{Grok 4.20 on the BenchLeader Index (method v1.4)}},
  year         = {2026},
  month        = nov,
  howpublished = {\url{https://www.benchleader.com/models/grok-4-20}},
  note         = {Data as of 2026-11-08}
}

Every badge style, the embeddable top-10 widget and how to cite the Index

Questions people ask

What are Grok 4.20's benchmark scores?
As of 8 Nov 2026, Grok 4.20 (thinking effort) scores 58.4 ±4.1 on the BenchLeader Index, strongest in instruction following (79) and weakest in agents & tools (50). That rests on independent benchmark results only, and every score on this page links to its publisher.
How does Grok 4.20 rank?
Grok 4.20 (thinking effort) is #124 of 593 ranked configurations on 8 Nov 2026. 230 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 69.4.
How much does Grok 4.20 cost per 1M tokens?
As of 8 Nov 2026, Grok 4.20 costs $1.25 per 1M input tokens and $2.50 per 1M output tokens, $1.56 blended (three input tokens per output token). The list price last changed on 18 Oct 2026.
Is Grok 4.20 better than Inkling 2 Small?
As of 8 Nov 2026, Grok 4.20 (thinking) and Inkling 2 Small are neck and neck on quality (58.4 vs 58.4 on the BenchLeader Index, inside the ±5.9 margin of error), and Inkling 2 Small costs 3.3× less, which makes it the better buy for most workloads. The head-to-head page has the benchmark-by-benchmark detail.
When was Grok 4.20 released?
SpaceXAI released Grok 4.20 on 17 Feb 2026. BenchLeader has results for it at 3 reasoning-effort settings.

Data as of 8 Nov 2026. Compare these configurations.

Cite as: BenchLeader, “Grok 4.20: benchmarks, pricing, speed and rank”, data as of 8 Nov 2026, https://www.benchleader.com/models/grok-4-20.