MDL
Qwen3.5 4B
Alibaba
Verified against Artificial Analysis · Jul 27, 2026
Capability
60
percentile index
Price /M
$0.03 / $0.15
in / out
Context
—
Avg score
47.6
8 benchmarks
AA Intelligence
20.1
index
Speed
30 tok/s
TTFT 0.5s
Released
Mar 2, 2026
About
Qwen3.5 4B is a open-weights model in the Qwen3.5 family from Alibaba. Benchmarked on 8 evals, averaging 47.6.
Benchmark scores · 8
avg 47.6 — higher is betterThe research behind this model
all papers mentioning it →Pair-In, Pair-Out: Latent Multi-Token Prediction for Efficient LLMsMay 26, 2026ExComm: Exploration-Stage Communication for Error-Resilient Agentic Test-Time ScalingMay 21, 2026VERDI: Single-Call Confidence Estimation for Verification-Based LLM Judges via Decomposed InferenceMay 11, 2026ReCrit: Transition-Aware Reinforcement Learning for Scientific Critic ReasoningMay 11, 2026
Mentions matched by name in title/abstract — from the arXiv + HF daily corpus.
Performance & price
Output speed30 tok/s
Latency (TTFT)0.5s
First answer1m 8s
Price in / out$0.03 / $0.15 per 1M
Blended price$0.06 / 1M
Coding index22.6
Median across API providers, via Artificial Analysis. Blended = 3:1 input:output. For reasoning models, latency includes thinking time. Methodology →
History
full ledger →Capability over time
Append-only ledger — every observed change to this model's numbers.
API providers
Artificial Analysis$0.15/M