Qwen 3.8 is Alibaba’s current flagship generation: Max opened at the top of the coding boards, Flash runs 1M-context agents at commodity prices, and the open-weight 3.8 family previews the Qwen 4 architecture.
Alibaba’s Qwen models are served via AlibabaCloud Model Studio and QwenCloud with OpenAI- and Anthropic-compatible endpoints. The 3.8 generation spans a 2.4T-parameter Max, fast Flash tiers and Apache-2.0 open dense models.
| Model | What it is for | Vendor list price |
|---|---|---|
| Qwen3.8-Max / -0902flagship · 2.4T MoE (95B active) · 1M ctx | Coding and agent flagship — the -0902 snapshot debuted at #1 on Code Arena WebDev (1,691) on 2 Sep 2026 | US$2 / US$6 per 1M (≈ RM8.16 / RM24.48) · cached US$0.25 |
| Qwen3.8-Flashproduction Flash · 1M ctx | Multimodal agent tier with built-in tools; OpenAI + Anthropic compatible | US$0.15 / US$0.47 per 1M (≈ RM0.61 / RM1.92) |
| Qwen3.8-Flash-Nextopen-weight preview · Qwen4 architecture | 125B (6B active + 51B N-gram) — first public look at Qwen 4 | ~US$0.12–0.20 / US$0.40–0.50 per 1M by provider |
| Qwen3.8-27BApache 2.0 open dense · 1M ctx | Self-hostable dense model for coding and office workloads | ~US$0.15–0.45 / US$2.00–3.20 per 1M by provider |
Last updated — Vendor list prices (Sep 2026), approx RM at 1 USD ≈ RM4.08 (spot, 19 Sep 2026). International (SG) pricing shown for Max; mainland China lists ¥12 / ¥36. Gateway prices for open variants vary by provider. Figures are vendor public list prices in USD per 1M tokens, standard tier — no batch, cache, promo or peak-hour discount applied. Vendors change list prices without notice: confirm on the vendor’s own pricing page before budgeting.
Discontinued or fully superseded models, kept for historical reference. Active models — even previous-generation ones still on sale — always stay in the lineup above.
| Model | Era & what it was for | Historic list price |
|---|---|---|
| Qwen3.7-Maxearly–mid 2026 | Predecessor flagship, later promo-priced at half | US$2.50 / US$7.50 per 1M (≈ RM10.20 / RM30.60) — historic list |
| Qwen3-Max2025 | The 2025 flagship generation | US$0.43 / US$1.72 per 1M (≈ RM1.75 / RM7.02) — historic list |
Recent announcements with outlet links (EN first, CN where available).
Scored 1,691 — three points past Claude Opus 5 Max and 22 above its own August checkpoint — at a blended ~US$5/M, the best score-to-price point on the board. TerminalBench 3.0 jumped 11.3 to 29.0. GPT-6 Astra Max has since taken #1 (1,800 on the 12 Sep board).
EN · TechTimes ↗EN · AlibabaCloud notice ↗EN · Arena Code WebDev board ↗125B hybrid (6B active + 51B N-gram) with QSA + GDN + Muon, extending 262K to 1M via YaRN.
EN · Qwen blog ↗EN · OrcaRouter ↗One partner for every digital layer – explore what else we build, host and grow for Malaysian businesses.
