Alibaba (Qwen)
Qwen 2.5 7B
model
qwen-2.5-7bBudgetStableFree★ Low cost in tierStreamingToolsBatchLong contextChina
Context
131K
Input / 1M tokens
$0.0700
Output / 1M tokens
$0.0700
About this model
Qwen 2.5 7B
Qwen 2.5 7B is the lightweight variant — fast, cheap, and capable for everyday tasks. Ideal as the small-model leg of a router that escalates harder queries to a larger model.
Released
2024-09
Training cutoff
2024-06
Parameters
7.6B (dense)
Best for
Use cases
- ✓High-volume classification, extraction, routing
- ✓Low-latency chat where 200ms TTFT matters
- ✓Edge deployment / on-device fallback
- ✓Cheap RAG retrieval+rerank pipelines
Strengths
What it does well
- ●Sub-second TTFT at typical concurrency
- ●Strong instruction-following despite size
- ●Tool-calling works reliably
- ●Cheap enough to run as a request-classifier in front of a larger model
Limitations
Know the trade-offs
- !Reasoning depth limited — not for multi-step math or planning
- !Long-context comprehension degrades past ~32K tokens in practice
- !Hallucinates more than the 72B on niche factual queries
Benchmarks
Published scores
Scores from official model cards and public leaderboards. Higher is better unless noted.
| Benchmark | Score | Measures |
|---|---|---|
| MMLU | 74.2 | General knowledge across 57 subjects |
| HumanEval | 57.9 | Python code generation, pass@1 |
| MATH | 49.8 | Mathematics, mixed difficulty |
Code example
Chat completion with Qwen 2.5 7B
from openai import OpenAI
client = OpenAI(
base_url="https://api.anoman.io/v1",
api_key="anm-sk-..."
)
response = client.chat.completions.create(
model="qwen-2.5-7b",
messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)Alternatives
Similar models in our catalog
Workloads that fit Qwen 2.5 7B often fit these too. Compare benchmarks and pricing before locking in a default.