Sglang beats vllm: • +7% on code gen tps

+14% on prose gen tps • -11% prefill tps setup: Qwen 3.8 27b + dflash2 @ 4x 3090 Anecdotally, its dominating vllm on ttft as well, need to bench that next. This uses A1RM4X_TV's recent PR (in thread), modified for official fp8

Trent Zock-RobbinsSep 105

AnyModel is advertising 5M free tokens for testing multiple AI models through one API—including GPT-6 Astra, Claude Fable 5, Kimi K3, and more.

How to get started: 1. Create an account: 2. Start the Telegram bot. 3. Claim the 5M-token offer. 4. Create

Tung AirSep 10

You can now use these models for FREE

Gemini 3.8 Flash • Muse Spark 1.3 • GLM 5.3 Flash • Gemini 3.1 Pro • DeepSeek V4 Flash • DeepSeek V4 Pro • GPT 5.6 Luna • Qwen 3.8 MAX Choose your free model here: Get your API key here

Simplifying AISep 1093

Qwen 3.8 Flash Next version :)

2x3090

GabuSep 981