Model comparison

DeepSeek-V3 vs Grok 4.20 (Non-Reasoning)

Grok 4.20 (Non-Reasoning) is the stronger model overall, scoring 48.6 to 39.5 on the Noometry Index. DeepSeek-V3 costs 3.9× less per token, which makes it the better buy when Grok 4.20 (Non-Reasoning)'s lead doesn't matter for your workload.

Last verified October 10, 2026. 26 shared benchmarks.

DeepSeek-V3

DeepSeek

39.5

Rank #166 Confirmed

Grok 4.20 (Non-Reasoning) xAI

48.6

Rank #54 Confirmed

Summary

They share 26 benchmarks with published results for both. DeepSeek-V3 scores higher in 1 category and Grok 4.20 (Non-Reasoning) in 7 categories; 7 gaps are clear of the uncertainty.
The widest gap is in reasoning, where Grok 4.20 (Non-Reasoning) leads 52.3 to 20.5.
The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 92.2% for Grok 4.20 (Non-Reasoning).
DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $1.25 / $2.50 for Grok 4.20 (Non-Reasoning).
Grok 4.20 (Non-Reasoning) accepts more context: 1M tokens versus 164K.
DeepSeek-V3 has downloadable open weights; the other is API-only.

Side by side

DeepSeek-V3 and Grok 4.20 (Non-Reasoning) specifications
	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
Provider	DeepSeek	xAI
Noometry Index	39.5	48.6
Released	2024-12-26	2026-02-17
Weights	Open	Proprietary
Context window	164K	1M
Max output	164K	30K
Input $ / M tokens	$0.24	$1.25
Output $ / M tokens	$0.90	$2.50
Results tracked	60	46

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding Too close to call

DeepSeek-V3: 42.3 (#106), Grok 4.20 (Non-Reasoning): 42.1 (#112)

Coding benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
WeirdML	36.1%	52.3%
LMArena Coding	1368	1459
Aider Polyglot	55.1%	—
LMArena WebDev	—	1375
SciCode	35.8%	—
BigCodeBench Instruct	50%	—
LiveBench Coding	70.9%	—
BigCodeBench Complete	62.2%	—
ALE-Bench	—	1,150
HumanEval+	86.6%	—
MBPP+	73%	—

Agentic & Tool Use Not comparable

DeepSeek-V3: —, Grok 4.20 (Non-Reasoning): 34.4 (#46)

Agentic & Tool Use benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
Terminal-Bench	—	57.3%
τ²-bench Banking	—	18%
LMArena Search	—	1189
METR Time Horizons	49.6%	—
Vending-Bench 2	—	4,663

Reasoning Grok 4.20 (Non-Reasoning) leads

DeepSeek-V3: 20.5 (#236), Grok 4.20 (Non-Reasoning): 52.3 (#32)

Reasoning benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
Kagi LLM Benchmark	52.3%	75%
LMArena Hard Prompts	1365	1451
DTBench	64.8%	90.1%
LMCA	15.5%	38.7%
Epoch Capabilities Index	135.94	151.98
ForecastBench	59.1	61.4
ARC-AGI-2	—	65.1%
SimpleBench	27.2%	—
NYT Connections (extended)	—	85.4%
ARC-AGI-1	—	89.5%
CritPt	0%	—
Chess Puzzles	—	24%
Thematic Generalization	—	63.8%
LiveBench Reasoning	65.8%	—
LiveBench Data Analysis	60.9%	—
BIG-Bench Hard	87.5%	—
HellaSwag	88.9%	—
LiveBench	66.9%	—
PIQA	84.7%	—
WinoGrande	85.2%	—

Math Grok 4.20 (Non-Reasoning) leads

DeepSeek-V3: 32.1 (#219), Grok 4.20 (Non-Reasoning): 48.2 (#65)

Math benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
OTIS Mock AIME 2024-2025	37.8%	92.2%
LMArena Math	1373	1455
FrontierMath (Tiers 1-3)	—	44.9%
FrontierMath Tier 4	—	17.1%
ProofBench	—	14%
Omni-MATH	40.3%	—
LiveBench Math	73.5%	—
MATH Level 5	75.5%	—
FrontierMath (Feb 2025 set)	1.7%	—

Knowledge Grok 4.20 (Non-Reasoning) leads

DeepSeek-V3: 37.5 (#155), Grok 4.20 (Non-Reasoning): 52.8 (#60)

Knowledge benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
GPQA Diamond	67.6%	89.3%
LMArena Expert	1351	1439
SimpleQA Verified	—	30.2%
MMLU-Pro	72.3%	—
Confabulations	26.1%	—
Vectara Hallucination Rate	6.1%	—
GPQA (HELM)	53.8%	—
ARC (AI2) Challenge	95.3%	—
MMLU	87.2%	—
TriviaQA	82.9%	—

Multimodal Not comparable

DeepSeek-V3: —, Grok 4.20 (Non-Reasoning): 33.3 (#98)

Multimodal benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
LMArena Vision	—	1263
Blueprint-Bench 2	—	0%
LMArena Document	—	1416

Multilingual Grok 4.20 (Non-Reasoning) leads

DeepSeek-V3: 48.5 (#143), Grok 4.20 (Non-Reasoning): 54.5 (#40)

Multilingual benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
LMArena Non-English	1358	1441
LMArena Chinese	1391	1481
LMArena French	1385	1476
LMArena German	1374	1465
LMArena Japanese	1333	1449
LMArena Korean	1319	1417
LMArena Russian	1373	1458
LMArena Spanish	1358	1443

Instruction Following Grok 4.20 (Non-Reasoning) leads

DeepSeek-V3: 72.8 (#130), Grok 4.20 (Non-Reasoning): 74.8 (#83)

Instruction Following benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
LMArena Instruction Following	1345	1420
LiveBench Instruction Following	81.5%	—
IFEval	83.2%	—

Long Context Grok 4.20 (Non-Reasoning) leads

DeepSeek-V3: 34.0 (#253), Grok 4.20 (Non-Reasoning): 45.5 (#34)

Long Context benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
LMArena Longer Query	1352	1437
Fiction.LiveBench	50%	—
CL-bench	—	22.2%
CL-bench Life	—	11.9%

Writing & Preference Grok 4.20 (Non-Reasoning) leads

DeepSeek-V3: 57.4 (#130), Grok 4.20 (Non-Reasoning): 65.7 (#44)

Writing & Preference benchmarks
Benchmark	DeepSeek-V3	Grok 4.20 (Non-Reasoning)
LMArena Text	1375	1451
LMArena Creative Writing	1364	1438
EQ-Bench Creative Writing	1472	1574
LMArena Multi-Turn	1389	1456
Short-Story Creative Writing	77%	—
WildBench	83%	—
LiveBench Language	49.1%	—

Frequently asked questions

Is DeepSeek-V3 better than Grok 4.20 (Non-Reasoning)?

Which is cheaper, DeepSeek-V3 or Grok 4.20 (Non-Reasoning)?

DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Grok 4.20 (Non-Reasoning) lists at $1.25 and $2.50.

Is DeepSeek-V3 or Grok 4.20 (Non-Reasoning) better for coding?

They score almost the same on coding (42.3 vs 42.1); test both on your own repository before choosing.

Which has the bigger context window?

Grok 4.20 (Non-Reasoning) does, with 1M tokens against 164K.

How many benchmarks do DeepSeek-V3 and Grok 4.20 (Non-Reasoning) share?

26 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Grok 4.20 (Non-Reasoning) has 46.