模型

GPT-5.6 Luna

The fast, low-cost tier of OpenAI's GPT-5.6 series, optimized for high-volume, latency-sensitive tasks such as classification, extraction, routing, and lightweight agentic steps. Approaches the larger GPT-5.6 tiers on many benchmarks while running several times faster at a fraction of the price.

Sarvam-30B

Sarvam AI's 30B-parameter Mixture-of-Experts reasoning model trained from scratch with only 2.4B active parameters per token. Optimized for real-time deployment and Indian languages, delivering strong reasoning, coding, and conversational performance while remaining efficient to serve. Open-weights.

Sarvam-105B

Sarvam AI's sovereign 105B-parameter Mixture-of-Experts model activating ~9B parameters per token, with a 128K-token context window. Trained on 12 trillion tokens across 22 Indian languages using 128 sparse experts with Multi-head Latent Attention and a custom low-fertility Indic tokenizer. Wins the majority of pairwise comparisons on Indian-language and STEM benchmarks.

GLM-5.2

Z.ai's (formerly Zhipu AI) flagship open-weight coding model with a 1M-token context window. Mixture-of-Experts architecture with 753B total parameters and ~40B active per request, featuring two cost-balancing reasoning modes. Tops several coding benchmarks while remaining a fraction of the cost of comparable proprietary frontier models. MIT-licensed weights.

Sarvam-M

Sarvam AI's 24B-parameter instruction-tuned model derived from Mistral-Small-3.1-24B, post-trained on English plus eleven major Indic languages (bn, hi, kn, gu, mr, ml, or, pa, ta, te). Delivers large relative gains on Indian-language, math, and programming benchmarks over its base model, with a hybrid reasoning mode for complex tasks.

Kimi K2.7 Code

Moonshot AI's latest open-source, coding-focused model in the Kimi K2 family, built to complete end-to-end programming tasks reliably over long contexts. A 1-trillion-parameter model that cuts reasoning token usage by roughly 30% versus K2.6 while improving coding and agent performance — +21.8% on Kimi Code Bench v2, +11.0% on Program Bench, and +31.5% on MLS Bench Lite for multi-language support. Released under a Modified MIT License and available via Kimi APIs and Hugging Face.

DiffusionGemma

Google DeepMind's experimental diffusion-based member of the Gemma 4 open model family. Unlike autoregressive models that generate text one token at a time, DiffusionGemma denoises a canvas of placeholder tokens to produce up to 256 tokens in parallel, finalizing output in one block. A Mixture-of-Experts model with 26B total parameters and 3.8B active per inference, delivering roughly 4x the throughput of similarly sized autoregressive Gemma models on local hardware. Excels at non-linear tasks like in-line editing, molecular sequencing, mathematical graphing, and self-correcting puzzles.

Nemotron 3 Ultra

NVIDIA's flagship open 550B-parameter Mixture-of-Experts model with 55B active parameters, built for frontier reasoning and orchestration in long-running agentic systems. Features hybrid Mamba-Transformer architecture, LatentMoE routing, multi-token prediction, and NVFP4 precision for 5x higher throughput. Achieves 30% lower cost-to-task-completion on agentic benchmarks. Supports 1M+ token context window with 95% accuracy on Ruler@1M.

Gemma 4 12B

Google's medium-size open-weight model with 12 billion parameters from the Gemma 4 family. Encoder-free unified multimodal architecture that natively processes text, image, audio, and video inputs without dedicated encoders. Features a 256K context window and supports 140+ languages. First medium-sized model capable of natively ingesting audio. Suitable for local deployment on GPUs with 16GB VRAM.

MiniMax M3

MiniMax's frontier open-weight model with 1M-token context window, native multimodality (text, image, video), and strong coding capabilities. Built on MiniMax Sparse Attention (MSA) architecture, achieving 59% on SWE-Bench Pro with significantly improved efficiency at long context.

Falcon-H1

TII's hybrid Mamba-Transformer model that outperforms comparable offerings from Meta's Llama and Alibaba's Qwen in the 30-70B parameter range. Designed for real-world AI on everyday devices and resource-limited settings with state-of-the-art efficiency.

DBRX

Databricks' open-source 132B parameter Mixture-of-Experts transformer model with 36B active parameters per input. Released under Databricks Open Model License, optimized for enterprise workloads including SQL generation and coding tasks.

Falcon 3 10B

TII's open-source 10B parameter model from the Falcon 3 family. Achieved number one position on Hugging Face's LLM leaderboard in its size category, outperforming Meta's Llama variants and other models under 13B parameters.

Ring-2.6-1T

InclusionAI's (Ant Group) trillion-parameter open-weights reasoning model with 63B active parameters per token. Built for real-world agent workflows with adaptive reasoning-effort modes. Features hybrid linear and MLA attention architecture with MIT license.

Yi-Lightning

01.AI's flagship large language model with enhanced Mixture-of-Experts architecture. Ranked 6th on Chatbot Arena with particularly strong results in Chinese, Math, Coding, and Hard Prompts categories. Features advanced expert segmentation and optimized KV-caching.

K2 Think

A 32 billion parameter open-weights reasoning model by LLM360/MBZUAI, built on Qwen2.5-32B. Trained with reinforcement learning and verifiable rewards for long chain-of-thought reasoning, agentic planning, and complex problem solving in math, science, and code.

Solar Pro 3

Upstage's powerful Mixture-of-Experts language model with 102B total parameters and 12B active parameters per forward pass. Optimized for Korean with strong English and Japanese support. Excels at complex reasoning, structured output generation, and agentic workflows.

Qwen 3.7 Plus

Alibaba's multimodal variant in the Qwen 3.7 family, optimized for vision understanding and multimodal tasks. Ranked

Gemini 3.5 Flash

Google DeepMind's balanced Gemini 3.5 model that pairs Pro-line reasoning quality with Flash-line latency and cost. Natively multimodal across text, image, audio, and video with a 1M-token context window, configurable thinking levels, and streaming function calling, tuned for high-throughput production workloads.

Gemini 3.1 Flash-Lite

Google's most cost-efficient Gemini model optimized for high-volume, low-latency use cases. Delivers 2.5x faster time to first token versus Gemini 2.5 Flash with full multimodal support. Ideal for agentic tasks, data extraction, translation, and classification.

Gemini 3 Flash

Google's balanced model combining Gemini 3 Pro's reasoning capabilities with the Flash line's latency, efficiency, and cost. Features configurable thinking levels, multimodal function responses, and streaming function calling for complex agentic workflows.

Granite 4.1 8B

IBM's dense decoder-only 8B parameter language model from the Granite 4.1 family. Supports 131K-token context, tool calling, RAG, code generation with fill-in-the-middle, text summarization, classification, and extraction across 12 languages. Released under Apache 2.0.

Granite 4.1 30B

IBM's largest dense decoder-only 30B parameter language model from the Granite 4.1 family. Trained on approximately 15T tokens with long-context extension up to 512K tokens. Supports tool calling, RAG, code generation, multilingual tasks across 12 languages. Released under Apache 2.0.

Laguna M.1

Poolside AI's flagship agentic coding model with 225B total parameters and 23B active (MoE). Trained from scratch in-house on 30T tokens across 6,144 NVIDIA Hopper GPUs. Optimized for complex multi-step software engineering tasks including codebase exploration, file editing, test running, and iterative debugging.