AI Models

Explore and compare the most powerful AI models available today

334 models tracked58 providers292 LMArena-rankedspec data verified through Nov 3, 2026

Quality vs Price

The highest LMArena Elo reached in each band of API input price (USD per million tokens). Only models with both a published price and an Elo are counted.

1480
Under $0.50MiMo V2.6 Pro95 models
1496
$0.50–1Gemini 3.8 Flash28 models
1528
$1–2Gemini 4 Carbon47 models
1509
$2–5Claude Opus 5.527 models
1504
$5–10Claude Fable 5.13 models
1448
$10+Claude Opus 4.15 models

Bars start at 1410 Elo so the gaps between bands are visible. Bar colour is the model maker's logo colour.

Gemini 4 Carbon

Google DeepMind · Coding

Our strongest model with state-of-the-art cybersecurity and long horizon capabilities.

Kling 4.0

Kling AI · Video Generation

Kuaishou’s Kling 4.0, fully available on the Kling API since 28 October 2026. It generates 3- to 30-second videos in a single run from text, images or video at 720p or 1080p, in 21:9, 16:9, 1:1, 9:16 or adaptive aspect ratios, with native audio and up to 10 keyframes. Its Omni reference endpoint accepts up to 15 inputs in a task, including up to 10 reference images and 5 reference videos, and can reference voices directly. With no input video it costs $0.157 a second at 720p and $0.235 at 1080p; with input video it costs $0.094 and $0.141 a second of total duration, which counts input and output. Kling lists 10-bit 4K as coming soon.

GPT-6.1 Terra

OpenAI · Coding

OpenAI has introduced GPT-6.1 Terra, a model designed for visual reasoning, computer use, and writing. According to OpenAI, standard API pricing for the model is $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens, which it states is 50% lower than prices for GPT-5.6 Terra and GPT-6.1 Sol. OpenAI stated that the model is available to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, as well as through the OpenAI API.

Kimi K3.1 Pro

Moonshot AI · Coding

Moonshot AI’s Kimi K3.1 Pro, released 24 October 2026: the first model in its K3.1 series, a post-trained version of Kimi K3 that Moonshot describes as its most capable 3T-class open-weight model, with a 1-million-token context window. Moonshot names coding, 3D creation and financial research as the main gains. It is on Kimi.ai, Kimi Work, Kimi Code and the Kimi API at $2 per million input tokens and $10 per million output tokens, with a Highspeed version at 2-3 times the speed for $4 and $20. Moonshot states the full weights will be released by 4 November 2026.

Spark X2.6

iFLYTEK · Coding

iFLYTEK’s Spark X2.6, unveiled on 22 October 2026 at the iFLYTEK 1024 Global Developer Festival in Beijing as a model with integrated on-device and cloud capabilities. Its open on-device version, Spark-X2.6-4B, keeps the architecture of Spark-X2.5-4B with improved post-training for coding, agents, long context, multilingual use and reasoning; it has three thinking levels (none, low, high), is published under Apache 2.0, and runs with a context length of up to 1,048,576 tokens.

FLUX 3 Image Klein

Black Forest Labs · Image Generation

Black Forest Labs' FLUX 3 Image Klein, released on October 15, 2026: a 5.6-billion-parameter version of FLUX 3 Image with open weights on Hugging Face under the FLUX Kommunity License. It supports text to image, image editing, up to ten reference images and bounding boxes, at up to 4K, and runs on consumer GPUs down to an RTX 3090. It comes in two variants: Distilled for fast generation and Base for fine-tuning and LoRA training. Through the BFL API it costs from $0.008 to $0.036 per image.

Nemotron 3.5 Bolt 74B-A9B

NVIDIA · Coding

NVIDIA’s Nemotron 3.5 Bolt 74B-A9B, released 15 October 2026: an open vision-language model with 74 billion total and 9 billion active parameters, compressed from Nemotron 3.5 Super VL. It takes images, video and text, returns text, and supports up to 256K tokens of context, with JSON output, tool calling and switchable reasoning. A 24GB version uses native 2-bit quantisation to fit in 24GB of memory and runs with llama.cpp, Ollama, LM Studio, Unsloth and vLLM. Weights are published under the OpenMDW License 1.1.

Nemotron 3.5 Super VL

NVIDIA · Coding

NVIDIA’s Nemotron 3.5 Super VL, released 15 October 2026: an open vision-language model with 120 billion total and 12 billion active parameters, combining Mamba-2, mixture-of-experts and attention layers with a C-RADIO vision encoder. It takes text, images and video, returns text, and supports up to 1M tokens of context. Weights are published under the OpenMDW License 1.1, on Hugging Face and as an NVIDIA NIM microservice.

Muse Spark Open

Meta · Coding

Muse Spark Open is the open-weight release of Muse Spark, Meta's agentic and coding model. It is a sparse mixture-of-experts transformer with 816 billion parameters in total and 42 billion active per token, with built-in perception encoders: it takes text, image, audio and video input and returns text, with a 1,048,576-token context window. Meta publishes it in two checkpoints - BF16 (about 1.65 TB of weights, served on eight GPUs such as 8 x B300) and NVFP4 (about 537 GB, served on one 4-GPU Blackwell node) - under the Muse Model License. It can be self-hosted with runtimes such as SGLang and vLLM, or called through hosted providers.

Muse Wonder

Meta · Coding

Muse Wonder is Meta's model announced on October 12, 2026, available through the Meta Model API (model ID muse-wonder-1.0) and Muse Code. It takes text, image, video, audio and PDF input, returns text, and has a 1,048,576-token context window. Meta lists API pricing of $4 per million input tokens and $20 per million output tokens. A larger version, Muse Wonder Full, is available only to Muse Code subscribers.

Vidu Q4 Preview

Shengshu Technology · Video Generation

Shengshu’s Vidu Q4 preview (viduq4-preview), added to the Vidu API on 7 October 2026 and described by Vidu as its next flagship for expressive audio and video. It generates 1- to 16-second clips at 24 fps from 540p up to 4K, with reference voices for consistent characters. At 4K it costs 78 credits per second.

Claude Haiku 5.5

Anthropic · Coding

Claude Haiku 5.5 is Anthropic's small model released on October 7, 2026 for high-volume, cost-sensitive tasks such as summaries, classification and extraction. Anthropic lists a 1M-token context window, 128K-token maximum output and a June 2026 knowledge cutoff, priced from $0.10 per million input tokens and $0.50 per million output tokens.

Showing 1–12 of 334 models

Frequently asked questions

Where do the scores on this site come from?+

Every score is somebody else’s published measurement, named where it appears. Rankings are LMArena Elo - a blind head-to-head human preference rating, shown with the date it was read. Benchmark percentages come from Epoch AI and from each benchmark’s own leaderboard. This site no longer publishes a score of its own: an editorial 0-100 index sat here until 13 September 2026 and was removed, because an opinion rendered as a number is hard to tell from a measurement.

Where does the pricing and spec data come from?+

Pricing, context windows and capability specs are read from each vendor's own documentation rather than aggregators. Every model page links the source it was read from and shows when the figures were last checked, so you can verify them yourself.

How do I compare two or more models?+

Use the compare icon on any model card to build a side-by-side comparison, or open one of the category comparisons under Comparisons. The spec table has presets (best value, cheapest, heavy workload) and a "differences only" view that hides rows where all selected models are identical.

Which AI model should I choose?+

It depends on your workload and budget, which is why this page plots quality against price. High-volume, cost-sensitive tasks usually favor cheaper models near the top-left of the value map; complex agentic or coding work tends to justify frontier-priced models. The editorial guide on each model page covers the trade-offs in detail.