Models

Google

Gemma 4 31B It

Released 2 Apr 2026

Google DeepMind's open-weights 30.7B dense multimodal model with a 256K-token context window and optional chain-of-thought thinking, instruction-tuned for long-context and multimodal tasks.

Google

Gemma 4 E2B It

Released 2 Apr 2026

A compact open-weights multimodal model from Google DeepMind, with a 128K token context window and optional built-in reasoning.

OpenAI

gpt-5.4-mini

Released 17 Mar 2026

OpenAI's efficient GPT-5.4 mini instruct model, built for coding and agentic workloads with optional reasoning depth and a 400K-token multimodal context window.

OpenAI

gpt-5.4-nano-2026-03-17

Released 17 Mar 2026

OpenAI's compact closed GPT-5.4 model for text and image input, with configurable reasoning effort and a 400,000-token context.

Mistral AI

mistral-small-2603

Released 16 Mar 2026

Open-weights 119B MoE model from Mistral AI (128 experts, 4 active, 6.5B active per token) unifying instruct, opt-in reasoning, and vision in a single 262K-context checkpoint served at an interactive latency class.

NVIDIA

NVIDIA-Nemotron-3-Super-120B-A12B

Released 11 Mar 2026

NVIDIA's open-weight text MoE model with optional reasoning, 12B active parameters and a long context window.

xAI

grok-4.20-0309-non-reasoning

Released 10 Mar 2026

xAI's non-reasoning Grok 4.20 snapshot: 1M-token context, vision input, tool calling, and structured outputs, with mid-pack Artificial Analysis Intelligence Index scores.

xAI

grok-4.20-0309-reasoning

Released 10 Mar 2026

xAI's chain-of-thought reasoning model in the Grok 4.20 generation, built for deliberate multi-step inference across text and image input.

xAI

grok-4.20-multi-agent-0309

Released 9 Mar 2026

xAI's multi-agent orchestration model: parallel sub-agent debate synthesized into one answer, built for deliberate reasoning over speed.

OpenAI

gpt-5.4

Released 5 Mar 2026

OpenAI's configurable chain-of-thought reasoning model with a 1M-token context window and native text, image, and file input.

OpenAI

gpt-5.4-pro

Released 5 Mar 2026

OpenAI's closed GPT-5.4 Pro reasoning model for complex analysis and very long text and image inputs, with configurable reasoning effort.

Qwen

qwen3.5-9b

Released 2 Mar 2026

A dense, open-weight 9.65B-parameter multimodal model from Qwen with a 262K-token native context window and text, image, and video input.

Qwen

qwen3.5-122b

Released 24 Feb 2026

Open-weight Qwen multimodal MoE model with about 10B active parameters, image input, optional extended thinking and up to 120K tokens through Stav.

Google

Gemini 3.1 Pro Preview

Released 19 Feb 2026

Google's frontier reasoning model with mandatory chain-of-thought inference, a 1M-token context window, and full multimodal input for long-horizon, complex tasks.

Anthropic

Claude Sonnet 4.6

Released 17 Feb 2026

A hybrid-reasoning Claude model from Anthropic with a 1M-token context window, selectable reasoning effort, and tool calling for coding, research, and agentic work.

Qwen

qwen3.5-397b-a17b

Released 16 Feb 2026

Qwen's open-weight multimodal MoE model with 397B total and 17B active parameters, image input and a 262,144-token context.

Anthropic

Claude Opus 4.6

Released 5 Feb 2026

One of Anthropic's Claude Opus 4 models — a one-million-token context window, optional adaptive thinking, and native document understanding for reasoning-heavy and agentic work.

Qwen

Qwen3-Coder-Next

Released 3 Feb 2026

Qwen's open-weight 80B/3B-active MoE coding model with a 262K context, tool calling and Apache-2.0 licence, suited to repository-scale code tasks and IDE assistants.

Google

Gemini 3 Flash Preview

Released 17 Dec 2025

Google's multimodal Flash model with a 1M+ token context and opt-in reasoning effort, built for fast Inference on complex, mixed-media tasks.

Mistral AI

ministral-14b-2512

Released 2 Dec 2025

Mistral AI's dense 14B multimodal Instruct model with Apache-2.0 open weights, a 262,144-token context window and native tool calling.

Mistral AI

ministral-3b-2512

Released 2 Dec 2025

Mistral AI's Apache-2.0 Ministral 3 3B Instruct model: text and image in, text out, 256k context, built for fast simple workloads.

Mistral AI

ministral-8b-2512

Released 2 Dec 2025

Mistral AI's dense 8B Instruct model with text and image input, a 262,144-token context window and open Apache-2.0 weights.

DeepSeek

deepseek-v3.2

Released 1 Dec 2025

DeepSeek's open-weight MoE model with optional reasoning, a 164K-token context window and solid tool use. Released under the MIT licence.

Anthropic

Claude Opus 4.5

Released 24 Nov 2025

Anthropic's frontier Claude Opus 4.5: a proprietary, text-and-image model built for complex coding, research, and long-document reasoning with optional extended thinking.

Anthropic

Claude Haiku 4.5

Released 15 Oct 2025

Anthropic's fastest Claude 4 model — built for high-throughput agentic and coding workloads, with a 200K-token context window, optional extended thinking, vision input, and strong safety benchmarks.

Anthropic

Claude Sonnet 4.5

Released 29 Sept 2025

Anthropic's instruct-tuned Claude Sonnet 4.5 — built for agentic coding, computer-use, and long-context reasoning via the Claude API.

Swiss AI Initiative

apertus-70b

Released 2 Sept 2025

A 70B-parameter, fully open-weight instruction model from the Swiss AI Initiative, built for transparency and multilingual coverage rather than peak benchmark performance.

NousResearch

Hermes-4-405B

Released 26 Aug 2025

Nous Research's open-weight 405B hybrid reasoning model based on Llama 3.1, with a 131K-token context and text input and output.

OpenAI

gpt-5-mini

Released 7 Aug 2025

OpenAI's compact GPT-5 variant with tunable reasoning effort, a 400K-token context window, and multimodal input for interactive document and image tasks.

OpenAI

gpt-5-nano-2025-08-07

Released 7 Aug 2025

OpenAI's fastest, lightest GPT-5 tier for high-volume classification, summarisation, and short completions — 400K context, vision and file input, mandatory reasoning traces.

OpenAI

gpt-oss-20b

Released 5 Aug 2025

OpenAI's open-weight 21B mixture-of-experts reasoning model under Apache 2.0, with adjustable reasoning effort and a 131K-token context.

Mistral AI

codestral-2508

Released 1 Aug 2025

Mistral AI's code-specialised API model for fill-in-the-middle completion, code correction and test generation, with a 256K-token context window.

Google

Gemini 2.5 Flash-Lite

Released 22 Jul 2025

Google's fastest, lowest-cost Gemini 2.5 model — a multimodal instruct model with a 1M-token context window, built for realtime, high-throughput Inference.

Qwen

qwen3-235b-a22b-instruct-2507

Released 21 Jul 2025

Qwen's open-weight Mixture-of-Experts model for instruction following, with a 262K-token context window. It answers directly, without a separate reasoning phase.

Mistral AI

Mistral-Small-3.2-24B-Instruct-2506

Released 20 Jun 2025

Deprecated 24B open-weights multimodal instruct model from Mistral AI with 131k context, tool calling, and strong document vision — use Small 4 for new work.

Google

Gemini 2.5 Pro

Released 17 Jun 2025

Google's closed-weight, reasoning-tuned flagship with a 1M-token context window, multimodal input, and native tool calling.

OpenAI

o3-pro-2025-06-10

Released 10 Jun 2025

OpenAI's higher-compute o3 reasoning model for deep, batch-style math, code, and research tasks via the Responses API.

Google

Gemini 2.5 Flash

Released 20 May 2025

Google DeepMind's efficiency-tier multimodal model with a 1M-token context window, optional chain-of-thought thinking, and native text, image, audio, video, and file input.

Barcelona Supercomputing Center

ALIA-40B Instruct (2601)

Released 22 Apr 2025

BSCs spanske og flerspråklige 40B-modell, instruct-varianten ALIA-40b-instruct-2601 (huggingface.co/BSC-LT/ALIA-40b-instruct-2601), Apache-2.0.

Meta AI

Llama-3.3-70B-Instruct

Released 6 Dec 2024

Meta's open-weights, dense 70B instruct model with a 128k-token context window, built for self-hosted or Sovereign Inference.

OpenAI

gpt-4-turbo

Released 9 Apr 2024

OpenAI's closed-weight GPT-4 Turbo brings a 128K-token context window and text+image input to complex, long-context tasks at interactive latency.

OpenAI

gpt-4

Released 14 Mar 2023

OpenAI's original GPT-4: a proprietary, text-only instruct model for professional writing, reasoning, and basic tool-calling workflows, with an 8k-token context window and a 2021 training cutoff.

OpenAI

gpt-3.5-turbo

Released 1 Mar 2023

OpenAI's GPT-3.5 Turbo is a fast, text-only chat model built for high-volume, simple conversational tasks via the Chat Completions API, with tool calling and JSON mode support.

OpenAI

gpt-oss-120b

OpenAI's open-weight, reasoning-tuned Mixture-of-Experts model — Apache-2.0 licensed, 131K context, self-hostable.

MiniMax

MiniMax M3.1 Flash

MiniMax's open-weights multimodal MoE model with a 1M-token context window and configurable reasoning effort.