MDL
Qwen3 4B 2507 Instruct
Alibaba
Verified against Artificial Analysis · Jul 13, 2026
Capability
46
percentile index
Price /M
$0 / $0
in / out
Context
—
Avg score
30.4
10 benchmarks
AA Intelligence
7.1
index
Weights
Open
Released
Aug 6, 2025
About
Qwen3 4B 2507 Instruct is a open-weights model in the Qwen3 family from Alibaba. Benchmarked on 10 evals, averaging 30.4.
Benchmark scores · 10
avg 30.4 — higher is betterThe research behind this model
all papers mentioning it →BlockPilot: Instance-Adaptive Policy Learning for Diffusion-based Speculative Decoding▲ 76 on HF · Jun 30, 2026From Trainee to Trainer: LLM-Designed Training Environment for RL with Multi-Agent Reasoning▲ 26 on HF · Jun 16, 2026KVpop -- Key-Value Cache Compression with Predictive Online Pruning▲ 20 on HF · Jul 6, 2026CausalMix: Data Mixture as Causal Inference for Language Model Training▲ 19 on HF · Jul 1, 2026EEVEE: Towards Test-time Prompt Learning in the Real World for Self-Improving Agents▲ 18 on HF · Jun 9, 2026
Mentions matched by name in title/abstract — from the arXiv + HF daily corpus.
History
full ledger →Capability over time
Append-only ledger — every observed change to this model's numbers.
API providers
Artificial Analysis$0/M