Model comparison

Claude Haiku 4.5 vs DeepSeek-V3

Claude Haiku 4.5 and DeepSeek-V3 score almost the same on the Noometry Index (39.5 vs 39.5), so choose on price, context window or the category you care about most.

Last verified October 10, 2026. 34 shared benchmarks.

Claude Haiku 4.5 Anthropic

39.5

Rank #165 Confirmed

DeepSeek-V3

DeepSeek

39.5

Rank #166 Confirmed

Summary

They share 34 benchmarks with published results for both. Claude Haiku 4.5 scores higher in 6 categories and DeepSeek-V3 in 2 categories; 6 gaps are clear of the uncertainty.
The widest gap is in math, where Claude Haiku 4.5 leads 44.9 to 32.1.
The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 66.7% for Claude Haiku 4.5 and 37.8% for DeepSeek-V3.
DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $1 / $5 for Claude Haiku 4.5.
Claude Haiku 4.5 accepts more context: 200K tokens versus 164K.
DeepSeek-V3 has downloadable open weights; the other is API-only.

Side by side

Claude Haiku 4.5 and DeepSeek-V3 specifications
	Claude Haiku 4.5	DeepSeek-V3
Provider	Anthropic	DeepSeek
Noometry Index	39.5	39.5
Released	2025-10-15	2024-12-26
Weights	Proprietary	Open
Context window	200K	164K
Max output	64K	164K
Input $ / M tokens	$1	$0.24
Output $ / M tokens	$5	$0.90
Results tracked	53	60

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

Category by category

Coding Claude Haiku 4.5 leads

Claude Haiku 4.5: 44.0 (#78), DeepSeek-V3: 42.3 (#106)

Coding benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
SciCode	43.3%	35.8%
WeirdML	45.4%	36.1%
LMArena Coding	1453	1368
SWE-bench Verified (bash only)	66.6%	—
Aider Polyglot	—	55.1%
LMArena WebDev	1330	—
SWE-bench Multilingual	64.7%	—
BigCodeBench Instruct	—	50%
LiveBench Coding	—	70.9%
BigCodeBench Complete	—	62.2%
ALE-Bench	653.48	—
HumanEval+	—	86.6%
MBPP+	—	73%

Agentic & Tool Use Not comparable

Claude Haiku 4.5: 33.6 (#52), DeepSeek-V3: —

Agentic & Tool Use benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
Terminal-Bench	35.5%	—
Berkeley Function Calling Leaderboard	68.7%	—
DeepResearch Bench	45.5%	—
BALROG	31.2%	—
ExploitBench	13.7%	—
METR Time Horizons	—	49.6%
Vending-Bench 2	458.89	—

Reasoning DeepSeek-V3 leads

Claude Haiku 4.5: 15.1 (#320), DeepSeek-V3: 20.5 (#236)

Reasoning benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
CritPt	0%	0%
LMArena Hard Prompts	1420	1365
DTBench	73.6%	64.8%
LMCA	30.9%	15.5%
Epoch Capabilities Index	142.41	135.94
ForecastBench	61.4	59.1
ARC-AGI-2	4%	—
SimpleBench	—	27.2%
Kagi LLM Benchmark	—	52.3%
NYT Connections (extended)	14.3%	—
ARC-AGI-1	47.7%	—
Chess Puzzles	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 Claude Haiku 4.5 leads

Claude Haiku 4.5: 44.9 (#78), DeepSeek-V3: 32.1 (#219)

Math benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
OTIS Mock AIME 2024-2025	66.7%	37.8%
Omni-MATH	56.1%	40.3%
LMArena Math	1396	1373
MATH Level 5	96.4%	75.5%
FrontierMath (Feb 2025 set)	5.9%	1.7%
LiveBench Math	—	73.5%
FrontierMath Tier 4 (v1)	2.1%	—

Knowledge Too close to call

Claude Haiku 4.5: 37.7 (#153), DeepSeek-V3: 37.5 (#155)

Knowledge benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
GPQA Diamond	71.2%	67.6%
MMLU-Pro	77.7%	72.3%
Vectara Hallucination Rate	9.8%	6.1%
GPQA (HELM)	60.5%	53.8%
LMArena Expert	1442	1351
SimpleQA Verified	13.2%	—
Confabulations	—	26.1%
ARC (AI2) Challenge	—	95.3%
MMLU	—	87.2%
TriviaQA	—	82.9%

Multimodal Not comparable

Claude Haiku 4.5: 26.8 (#118), DeepSeek-V3: —

Multimodal benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
Blueprint-Bench 2	0%	—
LMArena Document	1420	—

Multilingual Claude Haiku 4.5 leads

Claude Haiku 4.5: 49.9 (#129), DeepSeek-V3: 48.5 (#143)

Multilingual benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
LMArena Non-English	1377	1358
LMArena Chinese	1417	1391
LMArena French	1408	1385
LMArena German	1375	1374
LMArena Japanese	1339	1333
LMArena Korean	1347	1319
LMArena Russian	1381	1373
LMArena Spanish	1420	1358

Instruction Following DeepSeek-V3 leads

Claude Haiku 4.5: 71.4 (#149), DeepSeek-V3: 72.8 (#130)

Instruction Following benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
IFEval	80.1%	83.2%
LMArena Instruction Following	1414	1345
LiveBench Instruction Following	—	81.5%

Long Context Claude Haiku 4.5 leads

Claude Haiku 4.5: 43.6 (#92), DeepSeek-V3: 34.0 (#253)

Long Context benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
LMArena Longer Query	1427	1352
Fiction.LiveBench	—	50%

Writing & Preference Too close to call

Claude Haiku 4.5: 57.9 (#123), DeepSeek-V3: 57.4 (#130)

Writing & Preference benchmarks
Benchmark	Claude Haiku 4.5	DeepSeek-V3
LMArena Text	1396	1375
LMArena Creative Writing	1372	1364
WildBench	83.9%	83%
LMArena Multi-Turn	1409	1389
Short-Story Creative Writing	—	77%
EQ-Bench Creative Writing	—	1472
EQ-Bench 4	1064	—
LiveBench Language	—	49.1%

Frequently asked questions

Is Claude Haiku 4.5 better than DeepSeek-V3?

Claude Haiku 4.5 and DeepSeek-V3 score almost the same on the Noometry Index (39.5 vs 39.5), so choose on price, context window or the category you care about most.

Which is cheaper, Claude Haiku 4.5 or DeepSeek-V3?

DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Claude Haiku 4.5 lists at $1 and $5.

Is Claude Haiku 4.5 or DeepSeek-V3 better for coding?

Claude Haiku 4.5 scores higher on coding benchmarks: 44.0 versus 42.3 in the Noometry coding category.

Which has the bigger context window?

Claude Haiku 4.5 does, with 200K tokens against 164K.

How many benchmarks do Claude Haiku 4.5 and DeepSeek-V3 share?

34 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and DeepSeek-V3 has 60.