ModelsSep 9, 2026 · 03:528 sources
Inception Launches Mercury 2.5 LLM at 1,100 Tokens Per Second, Topping OpenRouter Speed Leaderboard
From huggingnews.com
Inception has launched Mercury 2.5, a language model that achieves 1,100 tokens per second on standard Nvidia GPUs through parallel token decoding, topping the OpenRouter speed leaderboard. The model, available on OpenRouter, Baseten, and the company's API, costs $0.04 per million input tokens and $0.15 per million output tokens, with input pricing 6x lower than Mercury 2. Benchmarks indicate a 40% increase in intelligence compared to the prior version, and it performs on par with Gemini 3.5 Flash and GPT 5.6 Luna Low without requiring specialized hardware.
Read the sourceSign in to suggest
Sources
+ 5 more
- SUPPORT@askalphaxiv“Together, these reduce global KV cache to 890 bytes/token, ~4x smaller than V4-Flash, and persistent KV by ~8x, while supporting 1M token contexts with much stronger overall performance.”x.com
- SUPPORT@teortaxestex“DeepSeek V4.1 Flash takes first place on AutomationBench-AA with 69%, equal to GPT-6 Astra (69%) and slightly above Grok 4.6 (67%).”x.com
- SUPPORT@ekzhang1“DeepSeek V4.1 has mathematically different outputs depending on whether you prefix cache hit”x.com
- SUPPORT@ollama“DeepSeek-V4.1-Flash is now rolling out for Pro plan subscribers.”x.com
- SUPPORT@therundownai“It beats GPT-5.6 Sol and Claude Opus 5 on several agentic, coding, and cyber benchmarks.”x.com
#inception#mercury#llm#openrouter#inference#speed#deepseek#benchmarkMarkdown
Content history (2)
- 2026-10-16 · Summary · vi · Replaced stored textMô hình Mercury 2.5 mới từ Inception có giá 0,04 USD cho mỗi triệu input token và 0,15 US…Mô hình Mercury 2.5 mới từ Inception có giá $0,04 trên mỗi triệu token đầu vào và $0,15 t…
- 2026-10-16 · Title · vi · Replaced stored textInception ra mắt LLM Mercury 2.5 với tốc độ 1.100 token/giây, dẫn đầu bảng xếp hạng tốc đ…Inception ra mắt Mercury 2.5 LLM với 1.100 token/giây, dẫn đầu bảng xếp hạng tốc độ OpenR…
One ranked edition, every morning
The stories that matter today. Free, cancel anytime.
Telegram @aidr_today