AI Daily · Oct 19, 2026: MiniMax M3.1 Flash launches with 1M context, free on multiple platforms
MiniMax M3.1 Flash (Space Bunny) launches with 1M context at $0.15/$0.60 per million tokens and is free for a limited time on OpenCode and Command Code; DeepSeek V4.1 Flash-1017 upgrades with SiliconFlow time-based pricing from $0.15.
Showing the latest briefing (Oct 19, 2026). Every day is generated the next morning.
Pick a date
-
MiniMax M3.1 Flash launches with 1M context and low pricing
MiniMax released M3.1 Flash (aka Space Bunny), an earlier version of which topped global leaderboards with 200T+ tokens. It supports a 1M-token context window and up to 512K output tokens, priced at $0.15 per million input tokens and $0.60 per million output tokens, and is available via API and web chat.
-
MiniMax M3.1 Flash free for limited time on OpenCode and Command Code
MiniMax M3.1 Flash (Space Bunny) is available on OpenCode free for the next 3 days, after ranking #1 there for 12 consecutive days with 106T tokens served. It is also free for a limited time in Command Code across all plans and subscribers, with a 1M-token context window.
-
DeepSeek V4.1 Flash upgraded to 1017 version with 763B parameters
The B.AI platform upgraded DeepSeek V4.1 Flash to V4.1 Flash-1017 without changing the model ID; the open-source 763B model supports 1M context and interleaved text-image input at unchanged pricing, with 11 benchmarks improved and DeepSWE v1.1 top-tier pass rate reaching 77.0%.
-
SiliconFlow sets DeepSeek-V4.1-Flash-1017 peak and off-peak token prices
Per 1M tokens on SiliconFlow, off-peak (14:00–24:00 UTC) is $0.15 input, $0.60 output, $0.003 cached input; peak (00:00–14:00 UTC) is $0.30 input, $1.20 output, $0.006 cached input.
-
DeepSeek V4.1-Flash-1017 free in Qoder until Oct 24
Qoder offers DeepSeek V4.1-Flash-1017 free for both free and paid users at 0.0x credits until Oct 24 23:59 UTC+8, after which usage returns to 0.1x.
-
MiniMax M3.1 Flash launches on GMI Cloud and Qoder
MiniMax M3.1 Flash, nicknamed Space Bunny, is now available on GMI Cloud and from day one in the Qoder suite; GMI cites 70% lower agentic task costs, 73.7% on DeepSWE v1.1 and 82% on CyberGym.
-
Command Code adds IDE extensions and several new models
Command Code shipped VS Code and Cursor IDE extensions and added DeepSeek V4.1 Flash, GLM 5.3 Flash and Ling 3.1 Flash with $80-$100 usage for $10, plus free stealth models Cat Jam and Beam 1.0, Gemini 4 Argon and Muse Wonder.
-
OpenAI restores OpenCode users after attack and refunds credits
OpenAI's team worked with OpenCode so users who are also OpenAI API customers are back in business, refunding credits that would have been drawn in the attack; Daybreak Blue access is available for a cyber-permissive defense setup.
-
Tencent Hunyuan open-sources Hy-Memory-30B-A3B memory model
Tencent Hunyuan released Hy-Memory-30B-A3B, an open model dedicated to memory management, and said more open-source releases are coming.
-
Tencent open-sources Octop 1.0.2 self-hosted AI workspace with team mode
Tencent released Octop 1.0.2, an open-source self-hosted AI workspace giving each person separate agents, memory and workspace, plus a team mode where one outcome description is split across multiple agents.