Qwen: Qwen3 VL 8B ThinkingReasoning

Best price — Output ($/1M tokens)

One row per access path (venue → underlying provider). Prices are each venue's published list price; performance is measured by Artificial Analysis. Your actual bill may vary by routing settings, fallbacks, region / ZDR, long-context tiers, and request features (JSON mode, tools, vision).

• #

  1

  Provider

  alibaba-cloud

  Class

  Official

  $ / 1M tokens

  Input $ / 1M tokens

  $0.18

  Output $ / 1M tokens

  $2.1

  Cache read $ / 1M tokens

  —

  +0.0% vs consensus

  TTFT

  2.3s

  reasoning 2.3s

  Throughput

  114.8 tps

  Uptime 24h

  —

  Requests 24h

  —

• #

  2

  Provider

  openrouter → qwen

  Class

  Reseller

  $ / 1M tokens

  Input $ / 1M tokens

  $0.18

  Output $ / 1M tokens

  $2.1

  Cache read $ / 1M tokens

  —

  +0.0% vs consensus

  TTFT

  —

  Throughput

  —

  Uptime 24h

  —

  Requests 24h

  —

• #

  3

  Provider

  deepinfra

  Class

  Direct host

  Disabled by deepinfra · disabled · fp8
  $ / 1M tokens

  Input $ / 1M tokens

  $0.18

  Output $ / 1M tokens

  $2.09

  Cache read $ / 1M tokens

  —

  -0.5% vs consensus

  TTFT

  —

  Throughput

  —

  Uptime 24h

  —

  Requests 24h

  —

• #

  4

  Provider

  zeroeval → deepinfra

  Class

  Reseller

  Disabled by zeroeval · disabled · fp8
  $ / 1M tokens

  Input $ / 1M tokens

  $0.18

  Output $ / 1M tokens

  $2.09

  Cache read $ / 1M tokens

  —

  -0.5% vs consensus

  TTFT

  —

  Throughput

  —

  Uptime 24h

  —

  Requests 24h

  —