China AI Hub
Comparisons / DeepSeek-V4-Pro vs Qwen3.8-Max: Price and Reasoning
DeepSeek-V4-Pro vs Qwen3.8-Max: Price and Reasoning
Updated: 2026-09-22
Compared entities
Comparison dimensions
Decision rule
DeepSeek-V4-Pro is the default for teams that value licence portability: MIT weights, $0.66/$1.98, and self-hosting as an exit. Qwen3.8-Max is the choice when you need Alibaba's regional endpoints, managed SLAs and enterprise contracting, and can absorb 3x the input price. Both list 1M context; the difference is who operates the model and what happens when you want to leave.
All figures below are from the China AI Hub database, last verified 2026-09-22. Scores and capabilities are labeled by source type; where a field is not publicly disclosed, we say so rather than estimating.
At a glance
| Dimension | DeepSeek-V4-Pro | Qwen3.8-Max |
|---|---|---|
| Status (database) | deprecated | active |
| Context window | 1,048,576 tokens | 1,048,576 tokens |
| Maximum output | 393,216 tokens | 131,072 tokens |
| Input price (per 1M) | $0.66 | $2.00 |
| Output price (per 1M) | $1.98 | $6.00 |
| Open weights | Yes (MIT) | No (proprietary) |
| API available | Yes | Yes |
| Reasoning | Yes | Yes |
| Coding | Yes | Yes |
| Vision / Video | No | Yes / Yes |
| Tool calling | Yes | Yes |
| Structured output | Yes | Yes |
| Benchmark records in DB | 4 | 5 |
Pricing
DeepSeek-V4-Pro lists input at $0.66 and output at $1.98 per 1M tokens; Qwen3.8-Max lists $2.00 input and $6.00 output — roughly three times the input price and three times the output price of the DeepSeek model. Both providers list their prices on official pages linked above. Price-per-token is only one axis: a model that needs fewer output tokens per task can cost less even at a higher per-token price.
Context and output
Both list a 1,048,576-token context window. They differ on maximum output: 393,216 tokens listed for DeepSeek-V4-Pro versus 131,072 for Qwen3.8-Max. For very long generations (whole-document writing, long code synthesis), the listed output ceiling differs materially.
Capabilities
Both list reasoning, coding, tool calling and structured output. The difference is in media: Qwen3.8-Max lists vision and video input; DeepSeek-V4-Pro lists neither. For document-image, chart and video workflows, only one of the two lists the capability.
Openness and deployment
DeepSeek-V4-Pro is open weight under MIT; Qwen3.8-Max is proprietary with API access only. Self-hosting is an option for the DeepSeek model and not for Qwen3.8-Max. Note the database lists DeepSeek-V4-Pro’s status as deprecated (preview 2026-04-24, GA 2026-08-13) — check the provider’s current lineup before building on it.
Benchmarks and verification
The database holds 4 benchmark records for DeepSeek-V4-Pro and 5 for Qwen3.8-Max, all labeled with source type (vendor-reported vs independent) and version. Different benchmarks cover different models, so counts are not comparable scores. We do not rank models on this page; the tables above state measured and vendor-listed facts, and the trade-offs below explain what to weigh.
Benchmark comparability is limited: the two models’ scores come from different test configurations — different benchmark versions, evaluation harnesses and tool setups — and several benchmarks cover only one of the two. Do not rank on headline scores alone.
Why each difference matters
China AI Hub analysis indicates the following per-dimension implications, drawn from the listed facts above.
- Cost: DeepSeek-V4-Pro has an advantage at ~3x lower input and ~3x lower output price; output-token-heavy reasoning workloads amplify that gap.
- Long-context: both list a 1M-token input window; DeepSeek-V4-Pro has an advantage in long single generations because it lists a 3x larger maximum output (393,216 vs 131,072 tokens).
- Reasoning: both list reasoning, so neither has a listed advantage on this axis.
- Coding: both list coding, tool calling and structured output; DeepSeek-V4-Pro’s price is more relevant for high-volume code generation, while Qwen3.8-Max adds vision/video for multimodal coding.
- Vision / media: Qwen3.8-Max is more relevant when document-image, chart or video input is required — it lists vision and video where DeepSeek-V4-Pro lists neither.
- Deployment: DeepSeek-V4-Pro has an advantage for self-hosting and data-residency (MIT open-weight, self-hostable); Qwen3.8-Max has a limitation here (proprietary, API-only).
- API: both expose an API; DeepSeek-V4-Pro is listed deprecated, which is the operational caveat to verify.
Trade-off summary
- Price: DeepSeek-V4-Pro lists substantially lower input/output prices.
- Media: Qwen3.8-Max lists vision and video; DeepSeek-V4-Pro lists neither.
- Output length: DeepSeek-V4-Pro lists a 3x longer maximum output.
- Openness: DeepSeek-V4-Pro is MIT open weight and self-hostable; Qwen3.8-Max is API-only.
- Status: DeepSeek-V4-Pro is listed as deprecated; Qwen3.8-Max as active.
The right choice depends on the workload — vision requirements, self-hosting needs, price sensitivity and how much output a task generates. Verify prices on the official pages before committing, as both change frequently.
Decision context
China AI Hub analysis indicates the following decision-context implications, drawn from the listed facts above.
For API developers. DeepSeek-V4-Pro lists ~3x lower input and output prices, and output-token-heavy reasoning amplifies that gap. Both list tool calling and structured output; Qwen3.8-Max adds vision/video for media input. DeepSeek-V4-Pro is listed deprecated. Latency is not publicly documented on this page.
For self-hosting. This is the clearest split on the page: DeepSeek-V4-Pro is MIT open weight and self-hostable; Qwen3.8-Max is proprietary with API access only. Hardware requirements and quantization are Not publicly documented on this page.
For coding agents. Both list coding, tool calling and structured output. DeepSeek-V4-Pro’s ~3x lower price is more relevant for high-volume code generation; Qwen3.8-Max adds vision/video for multimodal coding. Both list a 1M-token context. SWE-benchmark and terminal scores are not broken out (aggregate counts 4 vs 5).
For enterprise. Region, SLA, data-residency and compliance terms are not publicly documented — confirm with the vendor. Deployment differs: DeepSeek-V4-Pro is self-hostable (MIT) with a data-residency option, while Qwen3.8-Max is API-only. DeepSeek-V4-Pro’s deprecated status is an operational caveat to verify.
What is uncertain
- Benchmark comparability is limited — different versions, harnesses and partial coverage.
- DeepSeek-V4-Pro is listed deprecated (preview 2026-04-24, GA 2026-08-13); verify the current lineup.
- No third-party benchmark evidence is recorded for either model.
Sources
- DeepSeek API pricing
- DeepSeek V4 Preview announcement
- Alibaba Cloud Model Studio — Qwen3.8-Max
- Alibaba Cloud Model Studio — Model pricing
Labels used above: Official fact (prices, context windows, capabilities and license terms from primary provider sources), Vendor-reported claim (benchmark scores), and China AI Hub analysis (the “why each difference matters” reasoning, introduced as analysis). No third-party benchmark evidence is currently recorded for these models.