AI Daily · Nov 16, 2026: Alibaba launches Qwen4-Max with 1M context, native vision and lower pricing
Alibaba's Qwen4-Max brings a 1M-token context, native vision and $1.50/M input pricing to OpenRouter and Vercel; Tencent's Hy4 flash undercuts it further with two weeks of free usage.
Pick a date
-
Alibaba releases Qwen4-Max with native vision and 1M-token context
Qwen4-Max uses a new MoE architecture with native vision and a 1M-token context window; Alibaba says it beats Qwen3.8-Max using up to ~50% fewer tokens and cutting average cost per task by over 60% in internal tests.
-
Qwen4-Max priced at $1.50/$4.50 per million tokens on OpenRouter
On OpenRouter, Qwen4-Max costs $1.50/M input and $4.50/M output, down from $2/$6 for Qwen3.8-Max, with cache hits at $0.075/M and 1M context supporting text, image and video input.
-
Tencent launches Hy4 flash at $0.07 input, $0.281 output per 1M tokens
Hy4 flash is priced at $0.003 per 1M cached tokens, $0.070 input and $0.281 output, with 150+ tokens per second and about 1/20 the cost of Hy4 on everyday tasks; through Nov 30 it is free on WorkBuddy with TokenHub API at half price, and OpenRouter support is coming.
-
Hy4 Flash free with near-unlimited usage for two weeks
OpenCode offers Hy4 Flash free with near-unlimited usage for the next two weeks; the model has a 1M token context, is multimodal, and supports zero data retention.
-
Qwen4-Max goes live on OpenRouter, OpenCode Go and Vercel AI Gateway
Qwen4-Max, the first model in the Qwen4 series aimed at coding and long-horizon agents, is now available on OpenRouter, OpenCode Go and Vercel AI Gateway; Vercel notes up to 384K output tokens.
-
Qwen4-Max scores 43 on Artificial Analysis Intelligence Index, 9 points above predecessor
At xhigh reasoning effort, Qwen4-Max scores 43 on the Artificial Analysis Intelligence Index, 9 points above Qwen3.8-Max and 3 behind Kimi K3.1 Pro (46), at roughly 11% lower cost per task.
-
MiniMax H3 image models launch with 20% API discount
MiniMax's open-weight H3 image generation and editing series is live, with MiniMax-H3-Image and the faster Highspeed variant both offered at 20% off standard API pricing. Both support up to 16 reference images and native 4K output up to 20MP.
-
Tencent open-sources LoopForge self-evolving delivery platform
Tencent released LoopForge, a self-hosted platform that splits plan, code, review and test into separate agents with script gates, turning a team's delivery process into SOPs coding agents can loop-run in the cloud.
-
Command Code adds free Inkling 2 Small model and raises GLM-5.3 usage
Command Code's weekly recap lists GLM-5.3 usage raised to $70 in GOAT, Muse Ember at 2x usage on GOAT, and Inkling 2 Small as a free model, alongside Gemini 4 Carbon and Claude Fable 5.2.
-
Qoder offers 80% off all plans for Qwen4-Max launch
Qoder is running a limited-time 80% discount across every product and plan for Qwen4-Max; users update the app and select Qwen4-Max in the model selector, which replaces Qwen4-Max-Preview.