XAI GROK·4.3 +0.3 OPEN GPT·5.5 +0.5 GOOG GEMINI·3 Pro +1.0 GOOG GEMINI·3 Flash +1.0 GOOG GEMINI·3 Nano +1.0 ANTH OPUS·4.7 +0.3 NOUS HERMES·5 +1.0 PERP SONAR·Pro 2 +1.0 ANTH SONNET·4.5 +0.1 DEEP (REASONING)·R2 +1.0 COHE COMMAND·R+ 3 +1.0 DEEP DEEPSEEK·V4 +1.0 ALIB QWEN·3 Coder +1.0 ANTH HAIKU·4 · MIST MISTRAL·Large 3 +1.0 ALIB QWEN·3 Max +1.0 META LLAMA·4 Behemoth · OPEN GPT·5.0 · XAI GROK·4 Heavy · 01AI YI·Large 2 · META LLAMA·4 Maverick · META LLAMA·4 Scout · OPEN O-SERIES·4-mini · MIST CODESTRAL·25 · XAI GROK·4.3 +0.3 OPEN GPT·5.5 +0.5 GOOG GEMINI·3 Pro +1.0 GOOG GEMINI·3 Flash +1.0 GOOG GEMINI·3 Nano +1.0 ANTH OPUS·4.7 +0.3 NOUS HERMES·5 +1.0 PERP SONAR·Pro 2 +1.0 ANTH SONNET·4.5 +0.1 DEEP (REASONING)·R2 +1.0 COHE COMMAND·R+ 3 +1.0 DEEP DEEPSEEK·V4 +1.0 ALIB QWEN·3 Coder +1.0 ANTH HAIKU·4 · MIST MISTRAL·Large 3 +1.0 ALIB QWEN·3 Max +1.0 META LLAMA·4 Behemoth · OPEN GPT·5.0 · XAI GROK·4 Heavy · 01AI YI·Large 2 · META LLAMA·4 Maverick · META LLAMA·4 Scout · OPEN O-SERIES·4-mini · MIST CODESTRAL·25 ·

cat ~/joe/models/registry.json

# a running watch on frontier LLMs. updated weekly via cron.

tracked 24
providers 12
updated 2026-06-07
source models.json
xAI

Grok 4.3

[current] ▲ +0.3
released 2026-05-15
ctx 256K
modality textvisiontool

Real-time X integration; agent improvements; faster tool loops.

OpenAI

GPT 5.5

[current] ▲ +0.5
released 2026-05-01
ctx 400K
modality textvisionaudiotool

Unified reasoning model; auto-router replaces o-series picker.

Google DeepMind

Gemini 3 Pro

[current] ▲ +1.0
released 2026-04-22
ctx 2M
modality textvisionaudiovideotool

2M token context; native video understanding; Deep Think mode.

Google DeepMind

Gemini 3 Flash

[current] ▲ +1.0
released 2026-04-22
ctx 1M
modality textvisionaudio

High-throughput tier; multimodal at low latency.

Google DeepMind

Gemini 3 Nano

[current] ▲ +1.0
released 2026-04-22
ctx 128K
modality textvision

On-device tier; ships with newer Pixel/Android.

Anthropic

Claude Opus 4.7

[current] ▲ +0.3
released 2026-04-15
ctx 1M
modality textvisiontool

Long-context reasoning improvements; better tool use under uncertainty; reduced sycophancy.

Nous Research

Hermes 5

[current] ▲ +1.0
released 2026-04-10
ctx 256K
modality texttool

Open-source agent model; trained on real tool-call traces.

Perplexity

Sonar Pro 2

[current] ▲ +1.0
released 2026-04-01
ctx 200K
modality texttool

Search-grounded; web-native answers; citations first-class.

Anthropic

Claude Sonnet 4.5

[current] ▲ +0.1
released 2026-03-20
ctx 1M
modality textvisiontool

Speed/cost-optimized; near-Opus reasoning on most benchmarks.

DeepSeek

DeepSeek-R (reasoning) R2

[current] ▲ +1.0
released 2026-03-15
ctx 128K
modality text

Chain-of-thought reasoning successor to R1; open weights.

Cohere

Command R+ 3

[current] ▲ +1.0
released 2026-03-05
ctx 256K
modality texttool

RAG-specialized; tool-use leader at this tier.

DeepSeek

DeepSeek V4

[current] ▲ +1.0
released 2026-02-28
ctx 128K
modality texttool

MoE architecture refinements; aggressive pricing remains.

Alibaba

Qwen 3 Coder

[current] ▲ +1.0
released 2026-02-20
ctx 256K
modality text

Agent-tuned coder model; open weights.

Anthropic

Claude Haiku 4

[current] · —
released 2026-02-10
ctx 500K
modality textvision

Sub-second latency; agent workflows; budget tier.

Mistral

Mistral Large 3

[current] ▲ +1.0
released 2026-01-22
ctx 256K
modality textvisiontool

Improved multilingual; function calling parity with frontier.

Alibaba

Qwen 3 Max

[current] ▲ +1.0
released 2026-01-10
ctx 1M
modality textvisionaudio

Massive multilingual; strong code; competitive with frontier closed models.

Meta

Llama 4 Behemoth

[current] · —
released 2025-12-15
ctx 1M
modality textvision

2T total params; teacher model; not open-weight (yet).

OpenAI

GPT 5.0

[stable] · —
released 2025-12-12
ctx 400K
modality textvisiontool

Initial GPT-5 launch; replaced o3/o4-mini routing.

xAI

Grok 4 Heavy

[stable] · —
released 2025-11-30
ctx 256K
modality textvisiontool

Multi-agent reasoning tier (premium plus).

01.AI

Yi Large 2

[current] · —
released 2025-11-12
ctx 200K
modality text

Bilingual leader (en/zh); strong reasoning at mid-tier price.

Meta

Llama 4 Maverick

[current] · —
released 2025-10-01
ctx 1M
modality textvision

MoE; 17B active / 400B total; instruction-tuned flagship open weights.

Meta

Llama 4 Scout

[current] · —
released 2025-10-01
ctx 10M
modality textvision

10M token context; MoE 17B active / 109B total.

OpenAI

o-series 4-mini

[legacy] · —
released 2025-09-18
ctx 200K
modality texttool

Reasoning-tuned mini; superseded by GPT-5 auto-router.

Mistral

Codestral 25

[current] · —
released 2025-08-14
ctx 256K
modality text

Code-specialized; 80+ languages; FIM optimized.

# this registry is editorial. corrections / additions: /contact. auto-refresh runs weekly.