Qwen vs Gemma: benchmarks and API costs

Qwen3.8-Flash-Next leads SciCode at 50.58% versus 45.49%. Gemma 4 31B has the lower illustrated request cost at $0.0018; the other model costs $0.0025.

Updated · 6 sources · Methodology

DimensionWinnerMarginEvidence
SciCode (%)Qwen3.8-Flash-Next5.09 ppQwen3.8-Flash-Next 50.58% / Gemma 4 31B 45.49%12
Output speed (tokens/s)Qwen3.8-Flash-Next19.9 tokens/sQwen3.8-Flash-Next 55.4 / Gemma 4 31B 35.5 tokens/s
AA time to first answer token (s)Qwen3.8-Flash-Next11.87 sQwen3.8-Flash-Next 38.17 s / Gemma 4 31B 50.04 s
Estimated request costGemma 4 31B$0.000710K + 1K: Qwen3.8-Flash-Next $0.0025 / Gemma 4 31B $0.001843

AA coding quality benchmarks

MetricQwen3.8-Flash-NextGemma 4 31B (Reasoning)
SciCode (%)50.58145.492

Performance

MetricQwen3.8-Flash-NextGemma 4 31B (Reasoning)
Output speed (tokens/s)55.435.5
AA time to first answer token (s)38.1750.04

Prices and request cost

MetricQwen3.8-Flash-Next · LyceumGemma 4 31B · Novita AI
Input / 1M tokens$0.23$0.144
Output / 1M tokens$0.53$0.44
Context window (tokens)262,1445262,1444
Max output (tokens)131,0726131,0724
API documentationNovita AI · Chat API
Cost example: 10,000 input + 1,000 billed output tokens
Qwen3.8-Flash-Next$0.0025
Gemma 4 31B$0.0018

Coding, reasoning, context and cost

Coding benchmark

  • AA SciCode: Qwen3.8-Flash-Next 50.58% / Gemma 4 31B 45.49%.12

Reasoning benchmark

  • The configured AA first-answer wait is 38.17 s for Qwen3.8-Flash-Next and 50.04 s for Gemma 4 31B.12

Context window (tokens)

  • Context size cannot separate these endpoints: both list 262,144 tokens.12

Estimated request cost

  • For a smaller token bill, Gemma is $0.0018 against Qwen’s $0.0025 in the published scenario.43

Methodology and sources

SciCode is AA’s scientific-coding evaluation, reported as a percentage; higher is better. It is separate from the 10,000-input-token performance test and is not an in-editor Cline evaluation. Quality and performance use the configurations shown beside the model names. A current AA v5 Intelligence Index is displayed only where both models have a published value. AA performance uses the 10,000-input-token benchmark at each model’s displayed configuration; reasoning settings vary by model. Output speed is the median output rate. Answer latency measures the first answer token after reasoning, not the initial streamed reasoning chunk. For non-streamed reasoning and non-reasoning models, AA reports the same first-chunk and first-answer event. AA estimates answer latency from initial response and reasoning duration; it is not a separately sampled median. These are reference deployments, not Novita endpoint guarantees. The uncached cost example bills 10,000 input and 1,000 output tokens at each named seller’s tariff; it is not the cost of the AA benchmark. Prices, context and output limits use the cited seller sources.

  1. Artificial Analysis · Qwen3.8-Flash-Next · quality and performance
  2. Artificial Analysis · Gemma 4 31B (Reasoning) · quality and performance
  3. Lyceum · Qwen3.8-Flash-Next · API pricing and limits
  4. Novita AI · Gemma 4 31B · endpoint pricing and limits
  5. Lyceum · Qwen3.8-Flash-Next · native context specification
  6. Lyceum · serverless request limits

Build with Novita Model APIs

Create an API key and connect your application.

Start BuildingExplore Models