Noometry

Model comparison

GPT-6 Astra vs Qwen2.5 72B Instruct

GPT-6 Astra is the stronger model overall, scoring 70.8 to 31.9 on the Noometry Index. Qwen2.5 72B Instruct costs 8.2× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.

Last verified . 24 shared benchmarks.

GPT-6 Astra OpenAI

70.8

Rank #1 Confirmed

Qwen2.5 72B Instruct Alibaba (Qwen)

31.9

Rank #267 Confirmed

Summary

  • They share 24 benchmarks with published results for both. GPT-6 Astra scores higher in 9 categories and Qwen2.5 72B Instruct in 0 categories; 9 gaps are clear of the uncertainty.
  • The widest gap is in math, where GPT-6 Astra leads 93.5 to 19.3.
  • The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 100% for GPT-6 Astra and 8.1% for Qwen2.5 72B Instruct.
  • Qwen2.5 72B Instruct is cheaper at $1.40 / $5.60 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
  • GPT-6 Astra accepts more context: 1.05M tokens versus 131K.
  • Qwen2.5 72B Instruct has downloadable open weights; the other is API-only.

Side by side

GPT-6 AstraQwen2.5 72B Instruct
ProviderOpenAIAlibaba (Qwen)
Noometry Index70.831.9
Released2026-09-032024-09
WeightsProprietaryOpen
Context window1.05M131K
Max output128K8K
Input $ / M tokens$10$1.40
Output $ / M tokens$50$5.60
Results tracked5643

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

Category by category

Coding GPT-6 Astra leads

GPT-6 Astra: 73.7 (#2), Qwen2.5 72B Instruct: 33.2 (#260)

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
WeirdML93.6%16%
LMArena Coding14871292
DeepSWE74.1%—
FrontierCode53.3%—
LMArena WebDev1786—
FrontierSWE65.5%—
SciCode56.5%—
GSO79.4%—
BigCodeBench Instruct—45.8%
MirrorCode46.7%—
BigCodeBench Complete—55.9%
ALE-Bench2,951—

Agentic & Tool Use GPT-6 Astra leads

GPT-6 Astra: 52.9 (#3), Qwen2.5 72B Instruct: 22.1 (#133)

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
BALROG68.3%16.2%
APEX-Agents64.7%—
Remote Labor Index20.8%—
TheAgentCompany—5.7%
GDP.pdf34.2%—
METR Time Horizons—35.8%
Vending-Bench 215,515—

Reasoning GPT-6 Astra leads

GPT-6 Astra: 85.1 (#1), Qwen2.5 72B Instruct: 22.3 (#199)

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
LMArena Hard Prompts14621271
DTBench97.3%62.9%
LMCA64.4%13.4%
Epoch Capabilities Index166.45129
ARC-AGI-295%—
NYT Connections (extended)98.1%—
ARC-AGI-198.5%—
CritPt31.7%—
Chess Puzzles72%—
EBR-Bench76.2%—
Mystery Game Puzzles84%—
Bench to the Future 30.14—
BIG-Bench Hard—79.8%
ForecastBench—57.5
HellaSwag—84.8%
PIQA—82.6%
WinoGrande—82.3%

Math GPT-6 Astra leads

GPT-6 Astra: 93.5 (#2), Qwen2.5 72B Instruct: 19.3 (#287)

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
OTIS Mock AIME 2024-2025100%8.1%
LMArena Math14651283
FrontierMath (Tiers 1-3)93.7%—
FrontierMath Tier 497.6%—
ProofBench99%—
Omni-MATH—33%
MATH Level 5—63.2%
FrontierMath Erdős2.9%—

Knowledge GPT-6 Astra leads

GPT-6 Astra: 75.3 (#1), Qwen2.5 72B Instruct: 27.0 (#253)

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
GPQA Diamond95.8%49.1%
LMArena Expert14831245
Humanity's Last Exam54.8%—
SimpleQA Verified75.6%—
MMLU-Pro—63.1%
Confabulations—19.1%
Vectara Hallucination Rate8.7%—
GPQA (HELM)—42.6%
ARC (AI2) Challenge—94.5%
MMLU—85.3%
TriviaQA—71.9%

Multimodal Not comparable

GPT-6 Astra: 55.0 (#3), Qwen2.5 72B Instruct: —

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
LMArena Vision1281—
Blueprint-Bench 249.7%—
Furniture Assembly80%—
LMArena Document1468—

Multilingual GPT-6 Astra leads

GPT-6 Astra: 53.7 (#61), Qwen2.5 72B Instruct: 41.0 (#213)

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
LMArena Non-English14301252
LMArena Chinese14841272
LMArena French14561280
LMArena German14401234
LMArena Japanese13791180
LMArena Korean14261188
LMArena Russian14361264
LMArena Spanish14071256

Instruction Following GPT-6 Astra leads

GPT-6 Astra: 76.3 (#44), Qwen2.5 72B Instruct: 65.5 (#221)

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
LMArena Instruction Following14501254
IFEval—80.6%

Long Context GPT-6 Astra leads

GPT-6 Astra: 44.5 (#62), Qwen2.5 72B Instruct: 38.9 (#188)

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
LMArena Longer Query14561282

Writing & Preference GPT-6 Astra leads

GPT-6 Astra: 75.3 (#7), Qwen2.5 72B Instruct: 46.7 (#215)

BenchmarkGPT-6 AstraQwen2.5 72B Instruct
LMArena Text14411269
LMArena Creative Writing14181221
LMArena Multi-Turn14481272
EQ-Bench Creative Writing2173—
WildBench—80.2%

Frequently asked questions

Is GPT-6 Astra better than Qwen2.5 72B Instruct?

GPT-6 Astra is the stronger model overall, scoring 70.8 to 31.9 on the Noometry Index. Qwen2.5 72B Instruct costs 8.2× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.

Which is cheaper, GPT-6 Astra or Qwen2.5 72B Instruct?

Qwen2.5 72B Instruct is cheaper. It lists at $1.40 per million input tokens and $5.60 per million output tokens; GPT-6 Astra lists at $10 and $50.

Is GPT-6 Astra or Qwen2.5 72B Instruct better for coding?

GPT-6 Astra scores higher on coding benchmarks: 73.7 versus 33.2 in the Noometry coding category.

Which has the bigger context window?

GPT-6 Astra does, with 1.05M tokens against 131K.

How many benchmarks do GPT-6 Astra and Qwen2.5 72B Instruct share?

24 benchmarks have published results for both models. GPT-6 Astra has 56 scored results on Noometry and Qwen2.5 72B Instruct has 43.

Related comparisons

Go deeper