Model comparison

Claude Fable 5 vs DeepSeek-V3.2-Exp

Claude Fable 5 is the stronger model overall, scoring 66.8 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 69× less per token, which makes it the better buy when Claude Fable 5's lead doesn't matter for your workload.

Last verified October 10, 2026. 35 shared benchmarks.

Claude Fable 5 Anthropic

66.8

Rank #5 Confirmed

DeepSeek-V3.2-Exp

DeepSeek

44.3

Rank #78 Confirmed

Summary

They share 35 benchmarks with published results for both. Claude Fable 5 scores higher in 8 categories and DeepSeek-V3.2-Exp in 1 category; 9 gaps are clear of the uncertainty.
The widest gap is in reasoning, where Claude Fable 5 leads 76.8 to 22.1.
The biggest single-benchmark swing is ProofBench: 95% for Claude Fable 5 and 8% for DeepSeek-V3.2-Exp.
DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $10 / $50 for Claude Fable 5.
Claude Fable 5 accepts more context: 1M tokens versus 164K.
DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.

Side by side

Claude Fable 5 and DeepSeek-V3.2-Exp specifications
	Claude Fable 5	DeepSeek-V3.2-Exp
Provider	Anthropic	DeepSeek
Noometry Index	66.8	44.3
Released	2026-06-07	2025-09-29
Weights	Proprietary	Open
Context window	1M	164K
Max output	128K	66K
Input $ / M tokens	$10	$0.26
Output $ / M tokens	$50	$0.38
Results tracked	62	49

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

Category by category

Coding Claude Fable 5 leads

Claude Fable 5: 70.6 (#4), DeepSeek-V3.2-Exp: 46.5 (#65)

Coding benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
LMArena WebDev	1625	1362
SciCode	61%	38.9%
WeirdML	91.9%	39.5%
LMArena Coding	1519	1454
DeepSWE	69.9%	—
FrontierCode	53.5%	—
SWE-bench Verified (bash only)	—	70%
Aider Polyglot	—	74.2%
SWE-bench Multilingual	—	59%
FrontierSWE	47%	—
GSO	78.4%	—
MirrorCode	63.9%	—
ALE-Bench	2,041	—

Agentic & Tool Use Claude Fable 5 leads

Claude Fable 5: 54.0 (#2), DeepSeek-V3.2-Exp: 32.7 (#59)

Agentic & Tool Use benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
APEX-Agents	63.6%	21.3%
Vending-Bench 2	5,680	1,034
Terminal-Bench	—	39.6%
Berkeley Function Calling Leaderboard	—	56.7%
Remote Labor Index	16.1%	—
TheAgentCompany	—	42.9%
τ²-bench Banking	39.7%	—
PostTrainBench	41.8%	—
GBAEval	74.5%	—
GDP.pdf	30%	—
LMArena Search	1230	—

Reasoning Claude Fable 5 leads

Claude Fable 5: 76.8 (#6), DeepSeek-V3.2-Exp: 22.1 (#208)

Reasoning benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
ARC-AGI-2	89.2%	4%
Kagi LLM Benchmark	91.4%	52.2%
NYT Connections (extended)	92.7%	36.7%
ARC-AGI-1	98.5%	57%
CritPt	28.6%	2.9%
Chess Puzzles	41%	14%
LMArena Hard Prompts	1508	1434
DTBench	98.4%	87.7%
LMCA	61.1%	29.1%
Epoch Capabilities Index	162.06	146.27
SimpleBench	81.9%	—
EnigmaEval	39.3%	—
Thematic Generalization	—	65%
EBR-Bench	39.5%	—
Mystery Game Puzzles	52%	—
Surface Evolver Bench	95%	—
Bench to the Future 3	0.13	—

Math Claude Fable 5 leads

Claude Fable 5: 88.5 (#5), DeepSeek-V3.2-Exp: 41.7 (#87)

Math benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
OTIS Mock AIME 2024-2025	100%	87.8%
ProofBench	95%	8%
LMArena Math	1519	1435
FrontierMath (Tiers 1-3)	87%	—
FrontierMath Tier 4	90.2%	—
MathArena Final-Answer Competitions	—	57.7%
FrontierMath (Feb 2025 set)	—	22.1%
FrontierMath Erdős	0%	—
FrontierMath Tier 4 (v1)	—	2.1%

Knowledge Claude Fable 5 leads

Claude Fable 5: 62.2 (#25), DeepSeek-V3.2-Exp: 51.7 (#66)

Knowledge benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
GPQA Diamond	85.9%	83.4%
LMArena Expert	1534	1436
SimpleQA Verified	70.7%	—
Vectara Hallucination Rate	—	5.3%

Multimodal Not comparable

Claude Fable 5: 45.3 (#17), DeepSeek-V3.2-Exp: —

Multimodal benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
LMArena Vision	1324	—
Blueprint-Bench 2	38.6%	—
Furniture Assembly	35.8%	—
LMArena Document	1496	—

Multilingual Claude Fable 5 leads

Claude Fable 5: 57.3 (#9), DeepSeek-V3.2-Exp: 52.2 (#90)

Multilingual benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
LMArena Non-English	1481	1409
LMArena Chinese	1543	1461
LMArena French	1505	1433
LMArena German	1486	1440
LMArena Japanese	1506	1374
LMArena Korean	1488	1371
LMArena Russian	1504	1424
LMArena Spanish	1498	1440

Instruction Following Claude Fable 5 leads

Claude Fable 5: 78.6 (#8), DeepSeek-V3.2-Exp: 74.5 (#93)

Instruction Following benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
LMArena Instruction Following	1502	1413

Long Context DeepSeek-V3.2-Exp leads

Claude Fable 5: 46.3 (#23), DeepSeek-V3.2-Exp: 47.6 (#16)

Long Context benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
LMArena Longer Query	1509	1428
Fiction.LiveBench	—	83.3%
CL-bench	—	13.2%
CL-bench Life	—	9.5%

Writing & Preference Claude Fable 5 leads

Claude Fable 5: 75.9 (#5), DeepSeek-V3.2-Exp: 62.4 (#77)

Writing & Preference benchmarks
Benchmark	Claude Fable 5	DeepSeek-V3.2-Exp
LMArena Text	1491	1425
LMArena Creative Writing	1494	1403
EQ-Bench Creative Writing	1943	1515
LMArena Multi-Turn	1504	1427
EQ-Bench 4	1340	—

Frequently asked questions

Is Claude Fable 5 better than DeepSeek-V3.2-Exp?

Which is cheaper, Claude Fable 5 or DeepSeek-V3.2-Exp?

DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Claude Fable 5 lists at $10 and $50.

Is Claude Fable 5 or DeepSeek-V3.2-Exp better for coding?

Claude Fable 5 scores higher on coding benchmarks: 70.6 versus 46.5 in the Noometry coding category.

Which has the bigger context window?

Claude Fable 5 does, with 1M tokens against 164K.

How many benchmarks do Claude Fable 5 and DeepSeek-V3.2-Exp share?

35 benchmarks have published results for both models. Claude Fable 5 has 62 scored results on Noometry and DeepSeek-V3.2-Exp has 49.