Qwen vs Claude: coding, reasoning and price

Choose Qwen3.8 Flash on Novita AI for the lowest cost: $0.00197 per example request versus $0.06000 for Claude Opus 5.5, about 30× less; choose Claude for coding and reasoning, where it leads the Qwen3.8-Flash-Next reference by 34.3 pp on Terminal-Bench 4.0 and 23.4 pp on HLE.

Updated · 5 sources · Methodology

DimensionWinnerMarginEvidence
Coding benchmarkClaude Opus 5.534.3 ppTerminal-Bench 4.0: Claude 59.6% vs Qwen3.8-Flash-Next 25.3%12
Reasoning benchmarkClaude Opus 5.523.4 ppHLE: Claude 61.4% vs Qwen3.8-Flash-Next 38.0%12
Output speed (tokens/s)Claude Opus 5.545.1 tokens/sClaude 96.8 vs Qwen3.8-Flash-Next 51.7 tokens/s12
Estimated request costQwen3.8 Flash$0.05803Qwen3.8 Flash $0.00197 vs Claude $0.0600034

Coding and reasoning benchmarks

MetricQwen3.8-Flash-NextClaude Opus 5.5 (Max)
AA · Terminal-Bench 4.01225.3%59.6%
AA · HLE1238.0%61.4%
AA · SciCode1250.6%66.9%

Performance

MetricQwen3.8-Flash-NextClaude Opus 5.5 (Max)
Output speed (tokens/s)1251.796.8
Time to first answer token (s)1240.8676.8

Prices and request cost

MetricQwen3.8 FlashClaude Opus 5.5
Input / 1M tokens34$0.15$4.00
Output / 1M tokens34$0.47$20.00
Context window (tokens)341,000,0001,000,000
Cost example: 10,000 input + 1,000 billed output tokens
Qwen3.8 Flash$0.00197
Claude Opus 5.5$0.06000

When each option fits

Qwen3.8 Flash

  • For the lowest token bill, Qwen3.8 Flash on Novita AI costs $0.15 input and $0.47 output per 1M tokens; the example request is $0.00197, about 30× less than Claude.34
  • For shorter waits before the answer starts, the Flash-Next reference’s AA median time to first answer token is 40.8 s versus Claude’s 676.8 s.12
  • Treat it as a cost tier, not a quality match: the Flash-Next reference trails Claude by 34.3 pp on Terminal-Bench 4.0 and 23.4 pp on HLE.12

Claude Opus 5.5

  • For coding agents, Claude’s 59.6% on Terminal-Bench 4.0 is more than double the Flash-Next reference’s 25.3%.12
  • For scientific coding, Claude’s SciCode lead is 16.3 pp, 66.9% versus 50.6%.12
  • Claude’s AA output speed of 96.8 tokens/s is 1.9× the Flash-Next reference’s 51.7 tokens/s.12

Methodology and sources

Alibaba describes Qwen3.8 Flash as the production version of Qwen3.8-Flash-Next. Prices and context here refer to Novita AI’s qwen/qwen3.8-flash API, while AA benchmarks, output speed and time to first answer token use the separately named Qwen3.8-Flash-Next reference with a 256,000-token tested context; they are not one measured endpoint.145

Coding and reasoning rows use the same AA evaluation versions: Terminal-Bench 4.0, HLE and SciCode. The available official reports do not provide this exact current model pair under a common setup. AA output speed and time to first answer token are model references, not Novita endpoint guarantees; the latter is AA’s median for its 10,000-input-token prompt and includes reasoning for both models. Prices and context windows are Novita AI’s listed values for Qwen3.8 Flash and Anthropic’s for Claude Opus 5.5; the uncached example uses 10,000 input and 1,000 billed output tokens.

Costs are estimates at listed rates for uncached synchronous requests, not measured invoices; billed output includes reasoning. Tax, retries, discounts and negotiated capacity are excluded. Novita AI publishes this page using public sources, without a private workload benchmark.

  1. Artificial Analysis · Qwen3.8-Flash-Next
  2. Artificial Analysis · Claude Opus 5.5 (Max, Default Fallback)
  3. Anthropic · Opus 5.5 pricing
  4. Novita AI · Qwen3.8 Flash pricing
  5. Alibaba Cloud · Flash-Next and production Flash relationship

Build with Novita Model APIs

Create an API key and connect your application.

Start BuildingExplore Models