Claude Haiku 5.5 Max vs Qwen3.8 Max 0902 compared

On the headline score, Anthropic's Claude Haiku 5.5 (Max) and Alibaba's Qwen3.8 Max (0902) are almost level. The sharpest difference is price: Claude Haiku 5.5 Max costs $0.50 per million output tokens, against $6.00 for Qwen3.8 Max 0902. Everything below is read from the benchmark, price and speed data in Franklin AI's catalogue; nothing here comes from a hands-on test.

Token pricing and a sample bill

Claude Haiku 5.5 Max lists $0.10 per million input tokens and $0.50 per million output tokens; Qwen3.8 Max 0902 lists $2.00 and $6.00. Qwen3.8 Max 0902 costs 12 times as much per output token.

For a sample workload of 1 million input tokens and 200,000 output tokens, Claude Haiku 5.5 Max would cost $0.20 and Qwen3.8 Max 0902 $3.20, a difference of $3.00. Scale it up to summarising 1,000 documents of 8,000 tokens into 400-token summaries and the bill is $1.00 for Claude Haiku 5.5 Max against $18.40 for Qwen3.8 Max 0902. Blended at three input tokens per output token, the effective rates are $0.20 for Claude Haiku 5.5 Max and $3.00 for Qwen3.8 Max 0902.

Which is newer

Qwen3.8 Max 0902 came first, on September 2, 2026; Claude Haiku 5.5 Max followed on October 7, 2026. That is close enough to count them as the same generation.

Waiting time

Claude Haiku 5.5 Max generates about 233 tokens per second and Qwen3.8 Max 0902 about 38 tokens per second. Claude Haiku 5.5 Max is ahead by 510% on throughput. Before any text appears, Claude Haiku 5.5 Max takes 4.6 minutes and Qwen3.8 Max 0902 3.2 seconds. The faster writer, Claude Haiku 5.5 Max, gives away too much at the start to win on time: counting the first-token wait, Qwen3.8 Max 0902 finishes first for any reply shorter than about 13,000 tokens. Waits that long are felt on every chat-style request, though they matter less for batch jobs.

Put together, a 2,000-token answer would take about 4.8 minutes from Claude Haiku 5.5 Max and 56 seconds from Qwen3.8 Max 0902, counting the first-token wait.

Benchmark scores

Claude Haiku 5.5 Max comes out ahead on the intelligence index, 36.5 to 33.7. That is a modest edge: real, but small enough that the right set of tasks could reverse it. In relative terms, Claude Haiku 5.5 Max scores 8% higher.

Among 109 active scored models, Claude Haiku 5.5 Max sits 37th and Qwen3.8 Max 0902 44th.

There are no coding or math index scores for either model in the catalogue, which keeps this comparison at the aggregate level.

Score per dollar

Dividing the blended price by the intelligence index gives a rough cost per point: $0.0055 for Claude Haiku 5.5 Max and $0.089 for Qwen3.8 Max 0902, per million tokens. With the higher score and the lower cost per point, Claude Haiku 5.5 Max is the better value per point on these numbers.