MODEL REGISTRY

Models

Every frontier AI model, tracked. Specs, pricing, benchmarks, and wire coverage from the JustSaid newsroom.

MODELS
11
tracked
LABS
7
organizations
OPEN
2
open-weights
CLOSED
9
proprietary
ACTIVE
11
in production
FILTER 11 models
OpenAI 3 models
GPT-5.5 CLOSED

OpenAI's latest flagship model. Largest context window in the GPT family at 1M tokens. Available in standard and Pro tiers.

CTX 1.0M IN $5 / OUT $30
GPT-4.1 CLOSED

OpenAI's latest flagship with 1M token context window. Optimized for coding, instruction following, and long-context tasks.

CTX 1.0M MMLU 90.4 IN $2 / OUT $8
GPT-4o CLOSED

OpenAI's multimodal flagship. Accepts text, image, and audio input; generates text and audio output. Replaced GPT-4 Turbo as the default model.

CTX 128K HumanEval 90.2 IN $2.5 / OUT $10
Anthropic 3 models
Claude Opus 4.7 CLOSED

Anthropic's latest and most capable flagship. Significantly cheaper than Opus 4 while matching or exceeding it across all benchmarks.

CTX 200K IN $5 / OUT $25
Claude Opus 4 CLOSED

Anthropic's most capable model. State-of-the-art on coding benchmarks, extended thinking, and agentic tasks. The first model to sustain peak performance on multi-hour agent workflows.

CTX 200K GPQA Diamond 74.9 IN $15 / OUT $75
Claude Sonnet 4 CLOSED

Anthropic's best balance of intelligence and speed. Matches Claude Opus 3.5 on most benchmarks at lower cost and latency.

CTX 200K MMLU 86.7 IN $3 / OUT $15
Meta AI 1 model
Llama 4 Maverick OPEN

Meta's mixture-of-experts open model. 128 experts with 17B active parameters per forward pass. Competitive with GPT-4o and Gemini 2.0 Flash.

CTX 1.0M MMLU 92 IN $0.2 / OUT $0.6
Google DeepMind 1 model
Gemini 2.5 Pro CLOSED

Google's most capable model with native thinking, 1M token context, and multimodal reasoning across text, code, images, audio, and video.

CTX 1.0M MATH 92 IN $1.25 / OUT $10
xAI 1 model
Grok-3 CLOSED

xAI's flagship model trained on the Colossus supercluster (200K H100s). Strong on math, science, and coding benchmarks.

CTX 131K MATH-500 95.8 IN $3 / OUT $15
DeepSeek 1 model
DeepSeek-V3 OPEN

DeepSeek's MoE flagship. 671B total parameters, 37B active. Trained for $5.5M on 14.8T tokens — a fraction of competing labs' budgets.

CTX 131K MATH-500 90.2 IN $0.14 / OUT $0.28
Mistral AI 1 model
Mistral Large CLOSED

Mistral's flagship model. Strong multilingual performance and function calling. Available via La Plateforme and major cloud providers.

CTX 131K HumanEval 92.1 IN $2 / OUT $6