Qoder Launches Renewal Bonus Credits for Personal Plans
Starting Oct 21 (UTC+8), Qoder awards bonus Credits on every Personal plan renewal: Lite +500, Pro +1,000, Pro+ +2,000, Ultra +3,000, plus double Credits on a f…
Global AI Digest
-
-
B.AI Web Chat Adds Model Search for 70+ AI Models
B.AI's Web Chat now includes a model search feature, letting users type a model name in the selection window to instantly locate one of 70+ available AI models…
-
Voxtral 3 Transcribes 1891 Eiffel Recording at 6.5% WER
A third-party test ran Mistral's new Voxtral transcription model on a noisy 1891 French phonograph recording of Gustave Eiffel, matching 100 of 107 words in INA…
-
OpenCode Sees Hundreds of Daily Credential-Phishing Attempts via OAuth Flows
dax, an OpenCode developer, says the project faces hundreds of distinct daily attacks, citing an example of a PlanetScale OAuth flow rigged to send credentials…
-
OpenAI Extends Prompt Caching to 12 Hours in API, Keeps Cache Indefinitely for Codex and Work
OpenAI now offers 12-hour prompt caching in the API, while logged-in Codex and Work users keep cached context indefinitely with no penalty for resuming old conv…
-
GMI Cloud Offers Free Unlimited MiniMax M3.1 Flash Access Until October 21
GMI Cloud is offering free, unlimited access to MiniMax M3.1 Flash with no prerequisites, available until October 21.
-
GMI Cloud Benchmarks EAGLE-4 Speculative Decoding on NVIDIA B300 for Qwen3.8-27B
GMI Cloud benchmarked EAGLE-4 speculative decoding on NVIDIA B300 against other speculators for Qwen3.8-27B, reporting 703 tok/s at batch, 2.7x faster than plai…
-
MiniMax M3.1 Flash Runs 1.5x Faster in Atomic Agent, Builds 3D Scene in 8 Minutes
A third-party test shows MiniMax M3.1 Flash running 1.5x faster in Atomic Agent, where a local Qwen 3.8 27B model planned a Halloween 3D scene and MiniMax handl…
-
LiteLLM Adds Cache-Aware Routing for Self-Hosted Models on Kubernetes
LiteLLM now supports cache-aware routing for self-hosted models on Kubernetes, keeping each agent session on the same GPU pod so vLLM can reuse its prompt cache…
-
GMI Cloud Launches EAGLE-4 Speculative Decoding on B300 Fleet
GMI Cloud is a launch partner for EAGLE-4 and has it live on its NVIDIA B300 fleet, reporting 2.7x faster decoding than plain decoding on Qwen3.8-27B.
