18 text
DeepSeek's cost-efficient V4.1 tier and the first model built on the Causal Encoder-Decoder (CED) architecture: a 552B MoE activating 8B parameters on input and 16B on output.
IBM's compact dense reasoning model for mathematics, code generation, and multilingual dialogue.
Z.ai's efficient multimodal model for fast coding and long-horizon agent tasks.
Open-weight dense 27B vision-language model from Qwen for long-running agent tasks.
The GA release of DeepSeek's open 1.6T-parameter MoE flagship for advanced reasoning and long-horizon agents.
NVIDIA's open 30B MoE built for high-throughput agentic workloads.
The GA release of DeepSeek's efficiency-optimized V4 Flash, re-post-trained for coding, reasoning, and agent workflows.
Z.ai's frontier reasoning model for long-context agents and software engineering.
MoonshotAI's coding-focused Kimi model for end-to-end software engineering.
MiniMax's multimodal foundation model for long-horizon agentic work.
MoonshotAI's next-generation multimodal model built for long-horizon coding and multi-agent orchestration.
Efficient Mixture-of-Experts Gemma 4 with only 4B active parameters per token.
Google's fourth-generation open dense model with native multimodal understanding.
Large hybrid reasoning model supporting thinking and non-thinking modes.
OpenAI's budget-friendly reasoning model — fast and surprisingly capable.
OpenAI's compact open-weights model for efficient inference.
Meta's efficient 70B model matching Llama 3.1 405B on key benchmarks.
Compact 8B model from Meta's Llama 3.1 series.