Qwen 2 72B Instruct
May 2024
Qwen 2 72B InstructCompare inon
Benchmarks of Qwen 2 72B Instruct16
BenchmarkByScoreRank
Math1
Writing14
Legacy · saturated or no longer reported1
No benchmark matches
Ranking?
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperGLM 5.3 FlashunstatedOpenwins 77%-34%$0.15 / $0.50Memory (4-bit): 44 GB → 193 GB+5
BetterGPT-5 Minihighwins 64%+90%$0.053
OpenGemma 4 31Bunstatedwins 76%-56%$0.10 / $0.34
CheaperHy3unstatedOpenwins 76%-36%$0.13 / $0.53Memory (4-bit): 44 GB → 179 GBCheaperGPT-6 Lunamaxwins 75%-45%$0.068Function calling only with none · Chat Completions+3ModelScoreCostPer taskWhat changes
GLM 5.3 FlashunstatedOpenwins 77%-34%$0.15 / $0.50Memory (4-bit): 44 GB → 193 GB+5Hy3unstatedOpenwins 76%-36%$0.13 / $0.53Memory (4-bit): 44 GB → 179 GBGemma 4 31BunstatedOpenwins 76%-56%$0.10 / $0.34GPT-6 Lunamaxwins 75%-45%$0.068Function calling only with none · Chat Completions+3Gemma 4 26B A4BunstatedOpenwins 73%-28%$0.15 / $0.60ModelScoreCostPer taskWhat changes
GPT-5 Minihighwins 64%+90%$0.053GPT-6 Lunamaxwins 75%-45%$0.068Function calling only with none · Chat Completions+3GPT-5.6 Lunaxhighwins 75%+24%$0.085GPT-5.6 Terraxhighwins 77%+1,141%$0.63Prompts with >272K input tokensGPT-6.1 Solmaxwins 79%+1,003%$0.72Remove unsupported parameters+5ModelScoreCostPer taskWhat changes
Gemma 4 31BunstatedOpenwins 76%-56%$0.10 / $0.34
Mistral Medium 3.5unstatedOpenwins 73%+728%$1.50 / $7.50Memory (4-bit): 44 GB → 77 GBGemma 4 26B A4BunstatedOpenwins 73%-28%$0.15 / $0.60Qwen3.8 27BunstatedOpenwins 69%+164%$0.42 / $2.55max_tokens range · QwenCloud+4Qwen 3.5 122B A10BunstatedOpenwins 69%+203%$0.40 / $3.20Memory (4-bit): 44 GB → 75 GBModelScoreCostPer taskWhat changes
Qwen 3.7 Max Previewwins 78%—$2.50 / $7.50Context: not verifiedQwen3.6 Max Previewunstatedwins 77%—$1.30 / $7.80Include the word JSON · QwenCloudQwen3.7 Plusunstatedwins 77%—$0.32 / $1.28Qwen3.6 Plusunstatedwins 74%—$0.50 / $3Qwen3.5 397B A17BunstatedOpenwins 74%—$0.60 / $3.60Memory (4-bit): 44 GB → 242 GBQwen3 Max (Preview)unstatedwins 73%—$0.87 / $3.48Qwen 3Openwins 73%——Context: not verified+1Qwen3 235B A22B Thinking 2507unstatedOpenwins 70%—$0.23 / $2.30Memory (4-bit): 44 GB → 141 GB+1All Alibaba (Qwen) models →Run it yourself
estimateFits on✗ 8 GB (phone, laptop)✗ 16 GB✗ 24 GB GPU✓ 48 GB GPU · compressed✓ 80 GB GPU · 8-bit
Licencetongyi-qianwencommercial use with conditions
History
2 versionsBenchmarks
all categoriesChat & Knowledge
Arena Text · overall1,262#173
Arena Text · Chinese1,239.9#150
Arena Text · English1,222#166
Arena Text · French1,170.4#138
Arena Text · German1,150.8#140
Arena Text · instruction following1,181#165
Arena Text · Japanese1,110.9#124
Arena Text · Korean1,082.7#134
Arena Text · non-English1,175.9#167
Arena Text · Russian1,168.7#167
Arena Text · Spanish1,168.8#145
MATH Level 539.1#24
Arena Text · overall · full cohort1,203.2#167
GPQA Diamond · Epoch40.8#113
Writing & Language
Arena Text · creative writing1,223#169
Arena Text · creative writing · full cohort1,180.7#165
FAQ
How much does Qwen 2 72B Instruct cost?›
$0.35 per 1M input tokens and $0.40 per 1M output tokens (cheapest host: DeepInfra).
Is Qwen 2 72B Instruct free to use?›
There is no free API; the cheapest paid price is $0.35 per 1M input tokens. The weights are open, so you can run it yourself.
How good is Qwen 2 72B Instruct, and what is it best at?›
#173 of 196 in Chat & Knowledge, #169 of 195 in Writing & Language.
Is Qwen 2 72B Instruct better than Mistral Small 3?›
On Arena Text · creative writing, Qwen 2 72B Instruct scores 1,223 vs 1,227 — about the same.
How can I use Qwen 2 72B Instruct?›
Through the API of DeepInfra.
What is the context window of Qwen 2 72B Instruct?›
33K tokens.
When was Qwen 2 72B Instruct released? Is there a newer version?›
Released 2024-05-28. It is the current version in its line.
Is Qwen 2 72B Instruct open source?›
Yes, the weights are open (tongyi-qianwen licence).
Can I run Qwen 2 72B Instruct locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Qwen 2 72B Instruct read images?›
No, it accepts text only.
What is Qwen 2 72B Instruct?›
Qwen 2 72B Instruct is an open-weights language model from Alibaba (Qwen), released May 2024.
What are the best alternatives to Qwen 2 72B Instruct?›
GLM 5.3 Flash (cheaper), GPT-5 Mini (better), Gemma 4 31B (open weights).