Monolith-1.0 Released with Open Weights
DrDatta_AIIMS · x · 2026-07-18
The cited post introduces **Monolith-1.0**, an open-weight model designed for frontier reasoning. It utilizes a **1.57T parameter MoE** architecture, activating **49.5B** parameters per token. It natively supports a **1 million token** context window and was scaled using a two-stage YaRN curriculum. Trained on **12,288 Ascend 910C NPUs** using **60T tokens**, it claims to achieve state-of-the-art performance among open-source models at the time of release.\n\nThe post also shares several benchmark scores: **GPQA Diamond 95.9%**, **MMLU-Pro 96.2%**, and **AIME 2025 90%+**. The model weights, tokenizer, and evaluation framework are all released under the **MIT License**, with no waitlist required, and an open chat interface is available for direct testing.
Related event: Monolith-1.0: 1.57T Parameter MoE Model Released(2 posts)→
More from Models
- OpenCodex turns OpenAI’s Codex harness into a multi-provider coding workflow — arrakis_ai · 2026-07-21
- Researchers debate whether GPT-OSS ever had a clear harm case — aiamblichus · 2026-07-21
- Kimi K3 hits 89.4% peak on software tasks while Fable 5 is slightly steadier — FinanceYF5 · 2026-07-21
- Kimi K3 leads on Go, but Fable 5 wins Python, JavaScript, TypeScript and Rust — FinanceYF5 · 2026-07-21
- Kimi K3 reaches 89.4% pass@4 and tops the benchmark over GPT-5.6 Sol — FinanceYF5 · 2026-07-21
- Kimi K3 and Fable 5 now look much closer than the old open-vs-closed gap — FinanceYF5 · 2026-07-21