Skip to content
Runs in:FranceMade in:China
OVH AI Endpoints (GRA)

Qwen3-Coder-30B-A3B-Instruct

Tokonomix Editorial Team·Reviewed by Mes Kalkan··
Section 01

Speed analysis

Latency measured across all benchmark runs. P50 (median) and P95 (95th percentile) give a realistic picture of response speed under normal and peak load.

P50 latency (median)P95 latency100 runs
59791915780236403150006-2707-22ms
Section 02

Quality scores

Evaluation results from judge-model scoring across diverse task categories. Scores reflect coherence, accuracy and instruction-following.

100
Coding
99
Multilingual
90
Creative
Section 03

Pricing history

Direct provider rates per million tokens, plus a typical-conversation cost estimate.

💰
API rates — Qwen3-Coder-30B-A3B-Instruct
$0.0700 per 1M input tokens
$0.2600 per 1M output tokens
≈ <$0.0001 per typical conversation (800 tokens)
Input vs output price (per 1M tokens)
per 1M input tokens$0.0700
per 1M output tokens$0.2600

Pricing over time

Input & output per 1M tokens · step-line = price changes

$0.0700

input / 1M

— stable

$0.2600

output / 1M

— stable

2026-06-142026-06-282026-07-12
Input
Output
Price change
⟳ synced weekly
Section 04

Tokens per second

Throughput in tokens per second, derived from measured P50 latency. Higher is better; fluctuations track provider-side load.

Throughput (tokens / s)456 / avg 1313
33348

Estimated from P50 latency × 200 output tokens — the absolute number depends on this assumption; the trend is what matters.

Section 05

Capabilities

ownedBy: Qwen
Section 06

Availability

Availability

No measurements yet

We haven't recorded enough API calls to show availability stats for this model. Data appears once the model starts receiving live traffic.

Section 07

Tokonomix benchmark verdicts

⚖️
Endorsed by 1 judge
Independent LLM judges evaluated this model on our weekly intelligence tests
claude-sonnet-4-593/100 · 42 runs
37 correct1 partial4 wrong88% accuracy
2026-07-12

Quality jumps 6.7 points to 97.6 with 86% latency improvement

Qwen3-Coder-30B-A3B-Instruct demonstrates significant improvements across both performance and quality metrics in this benchmark window. The overall quality score increased from 90.8 to 97.6, marking a 6.7-point gain that places the model in excellent territory. Latency saw dramatic optimization, with the p50 dropping from 30,286ms to 4,333ms, representing an 86% improvement that makes the service far more practical for real-time applications. Category performance shows continued strength in multilingual capabilities with a score of 99, up from 97. Reasoning performance reached a perfect 100, while coding maintained strong results at 94, up slightly from 92. The creative writing category was not evaluated in the current window, so trends there cannot be assessed. This represents a complete reversal from the previous period's quality decline, suggesting infrastructure or model optimizations have taken effect. Both test windows evaluated five runs, maintaining consistent testing methodology. Users can expect substantially faster responses without sacrificing output quality, making this an ideal moment to evaluate the service for latency-sensitive applications.

Quality

97.6

Latency p50

4,333 ms

Test runs

5

Quality up 6.7 points Latency improved 86% Perfect reasoning score achieved Multilingual performance strengthened
Last automated test
Jul 22, 2026 · 02:00 UTC · Speed benchmark
P50 latency
439 ms
P95 latency
469 ms
Errors
0 / 6 runs
Last reviewed by Tokonomix Team·July 22, 2026