Qwen 3.7 Max is Alibaba's best model for research — it scores 89/100 vs 88/100 for Qwen 3.8 Max, at $2.5/1M input tokens. Across all providers, Claude Fable 5 still leads research at 100/100 — worth considering if you're not committed to Alibaba.
The shortest way to see the safest default, the lower-cost option, and the specialist pick before you read deeper.
Qwen 3.7 Max is the safest overall answer here when you want the strongest default instead of the lowest list price.
Meta: Llama 3.1 8B Instruct is the lower-cost option to start with when you still need useful output at scale.
Qwen 3.8 Max is the better pick when large documents, transcripts, or knowledge-heavy work lead the decision.
Qwen 3.7 Max leads Alibaba's lineup for research at 89/100 ($2.5/1M input, 1M context).
Qwen 3.8 Flash is the value pick at $0.16/1M input with a research score of 78/100.
Claude Fable 5 (Anthropic) is the overall research leader at 100/100 if provider choice is open.
Choose Qwen 3.7 Max when research quality is the priority and you're staying on Alibaba.
Choose Qwen 3.8 Flash when token volume matters more than peak quality.
Teams open to other providers should also evaluate Claude Fable 5 before committing.
Switch the scoring lens to see whether the top answer changes when you care more about cost, speed, or long-document work.
Alibaba / Balanced / Aug 6, 2026
Best Chinese flagship — beats GPT-5.6 Sol on coding, #2 globally for vision.
Ranks models by the broadest mix of coding, writing, research, and long-context usefulness.
You need independently verified benchmarks or Western data residency.
The fastest way to see where the recommendation shifts when your priority changes.
Agent-first Qwen flagship, superseded by Qwen 3.8 Max.
Best Chinese flagship — beats GPT-5.6 Sol on coding, #2 globally for vision.
SWE-bench Pro 62.5 at sixteen cents per million input.
Every figure below is the provider's list price or a published capability score — the same numbers the recommendation on this page is built from.
| Model | Input | Output | Est. month | Context | Speed | Coding | Writing | Research |
|---|---|---|---|---|---|---|---|---|
| Qwen 3.7 MaxAlibaba | $2.50/1M | $7.50/1M | $40 | 1M tokens | Balanced | 89 | 83 | 89 |
| Qwen 3.8 MaxAlibaba | $2.00/1M | $6.00/1M | $32 | 1M tokens | Balanced | 93 | 85 | 88 |
| Qwen 3.8 FlashAlibaba | $0.16/1M | $0.47/1M | $2.54 | 991k tokens | Very fast | 84 | 76 | 78 |
Capability scores are out of 100 and reflect our own weighting of published benchmarks and production signals — see how we evaluate models. “Est. month” assumes 10M input and 2M output tokens at list price, with no batch or caching discounts applied, so treat it as a ceiling.
What each one is genuinely good at, where it falls down, and the situations we would steer you away from it — not just the headline score.
Alibaba's first closed-weight flagship — an agent-first model with native extended thinking, built to run autonomously for up to ~35 hours firing thousands of tool calls.
Starting fresh — Qwen 3.8 Max is stronger, cheaper, and multimodal.
A capable agent-first flagship that was superseded within three months by Qwen 3.8 Max at a lower price. Choose it only if its specific extended-thinking behavior fits your agent stack; otherwise 3.8 Max is the better Qwen.
Announced at Alibaba Cloud Summit May 20, 2026. API-only on DashScope/Model Studio. Cached input $0.25/1M.
Alibaba's largest model ever — a 2.4-trillion-parameter MoE (95B active) multimodal flagship that beat GPT-5.6 Sol on SWE-bench Pro and ranks #2 globally for vision.
You need independently verified benchmarks or Western data residency.
The strongest Chinese multimodal flagship and a legitimate SWE-bench Pro upset over GPT-5.6 Sol. If vision matters, only Fable 5-class models beat it — at 3–8x the price. Wait for independent evals before betting production on the self-reported numbers.
Announced August 3, 2026 on Alibaba Cloud Model Studio; open weights promised a week after launch. $2/$6 is first-party Model Studio pricing; cache reads from $0.17/1M. Announcement moved Alibaba stock +7% in Hong Kong.
Alibaba's preview of the Qwen4 architecture — 125B parameters with only 6B active per token, at sixteen cents per million input.
You need Alibaba's maximum capability — that is Qwen 3.8 Max — or a SWE-bench Verified number.
One of the best coding-score-per-dollar picks in the catalog. Route volume work here and reserve Qwen 3.8 Max or a frontier model for the hard cases.
Released August 26, 2026. The open-weight release is Qwen3.8-Flash-Next, a preview of the Qwen4 architecture: 125B mixture-of-experts with 6B active per token, a 51B n-gram embedding table and a 4B multi-token prediction layer. Qwen 3.8 Flash is the production API version on Qwen Cloud at $0.16/$0.47.
UseRightAI recommendations are based on practical decision factors people actually feel in day-to-day use.
Newsletter
Useful if you care about ranking shifts, pricing changes, or a better recommendation appearing in this decision path.
No spam. Useful updates only. Affiliate disclosures always clearly labeled.
Qwen 3.7 Max — it scores 89/100 on research in this directory, ahead of Qwen 3.8 Max at 88/100. Agent-first Qwen flagship, superseded by Qwen 3.8 Max.
Not overall. Claude Fable 5 (Anthropic) leads the directory for research at 100/100 vs Qwen 3.7 Max's 89/100. Qwen 3.7 Max is the best pick if you're staying within Alibaba's ecosystem.
Qwen 3.8 Flash at $0.16/1M input tokens (research score: 78/100). Use it for volume work and reserve Qwen 3.7 Max for the tasks where quality matters most.
$2.5/1M input tokens and $7.5/1M output tokens via the API. Context window: 1M tokens. On a moderate month — 10M input and 2M output tokens — that works out to about $40.00, against $2.54 for Qwen 3.8 Flash.
Behind the mid-2026 Western frontier on agentic coding, and now superseded by Qwen 3.8 Max — which is also cheaper. API-only with no open weights — a break from Qwen tradition that removes the self-hosting advantage. Concretely, avoid it if starting fresh — Qwen 3.8 Max is stronger, cheaper, and multimodal. If none of that is negotiable, Claude Fable 5 (Anthropic) is the cross-provider leader at 100/100.
marathon agent sessions — designed for ~35-hour autonomous runs with thousands of tool calls, graduate-level science reasoning (92.4 GPQA Diamond, near the top of the field), and long agent loops made cheap by the 90% cached-input discount ($0.25/1M cached). Its 1M-token context window is the practical limit on how much you can hand it in one go.
Qwen 3.7 Max scores 89/100 on research against 78/100 for Qwen 3.8 Flash, at 16x the input price. That premium is worth it on work where a wrong answer costs real time or money, and hard to justify on high-volume, low-stakes calls. Most teams run both and route by task rather than picking one.