Scores reflect fields captured in verified launch records and can change when better primary disclosures are added. They do not evaluate undisclosed internal practices.
Commercial API available to developers; enterprise deployment on-premises, in the cloud, or hybrid
—
—
—
commercial API and enterprise AI system
GPT-4b micro
OpenAI
3
50
—
developed for research purposes; not broadly available
64000
—
—
experimental research
GPT-5.2
OpenAI
3
50
—
In ChatGPT: GPT-5.2 Instant, Thinking, and Pro begin rolling out today to paid plans; In the API: available now to all developers.
262144
—
—
API and ChatGPT (model variants: GPT-5.2 Instant, GPT-5.2 Thinking, GPT-5.2 Pro)
GPT Image 1.5
OpenAI
2
33
—
Model rolling out in ChatGPT for all users and available in the API as GPT Image 1.5; the new Images experience is rolling out to most ChatGPT users, with Business and Enterprise access coming later; global.
—
—
—
API and ChatGPT rollout
Claude Mythos Preview
Anthropic
5
83
—
restricted access (restricted-access rollout)
1000000
25.000000
125.000000
Preview
Laguna XS.2
Poolside
3
50
Apache 2.0
Preview; Hugging Face, Poolside API, and OpenRouter
—
—
—
open weights
Claude Fable 5
Anthropic
4
67
—
Publicly accessible via API and Enterprise; included in subscription plans through 2026-06-22 after which access requires purchase of usage credits; some sensitive-topic access restricted and Mythos 5 limited to vetted Project Glasswing participants
—
10.000000
50.000000
Public release (Mythos-class); available via API and Enterprise
Claude Sonnet 5
Anthropic
4
67
—
All Claude plans and Claude API
—
2.000000
10.000000
general availability and API
Laguna XS 2.1
Poolside
6
100
OpenMDW-1.1
Poolside API, OpenRouter, Hugging Face weights, and local deployment
256000
0.100000
0.200000
open weights
GPT-Live
OpenAI
2
33
—
Rolling out to ChatGPT users globally (ChatGPT Voice); API availability planned/coming soon. (Update 2026-07-31: SynthID watermarking and API verification access added for supported audio.)
—
—
—
voice model (full-duplex)
GPT-5.6 (Sol, Terra, Luna)
OpenAI
2
33
—
General availability following a limited preview
—
—
—
general availability
Kimi K3
Moonshot
4
67
License requires companies operating a model-as-a-service business with > $20M total annual revenue (over 12 months) to enter a separate agreement with Moonshot before commercial use.
Weights published for anyone to download; model can be run locally on user hardware
1000000
—
—
open-weights release (weights published for public download)
Claude Opus 5
Anthropic
2
33
—
Available immediately; default on Claude Max and strongest model on Claude Pro
—
—
—
general availability
MiniMax H3
MiniMax
3
50
MiniMax H3 Community License Agreement
Hosted API globally; downloadable weights restricted to the license's applicable territories
—
—
—
API and open weights
abliterated-model-large-v2
Abliteration.ai
4
67
—
Commercial API access
—
5.000000
5.000000
API
Muse Glimmer
Meta
4
67
Apache 2.0
Consumer hardware with 24 GB or 32 GB of memory
131072
—
—
open weights
Muse Glimmer
Meta Platforms
2
33
—
Developers can download and modify the model
—
—
—
open weights
Nemotron 3.5 Lightning
Nvidia
4
67
OpenMDW-1.1
Available through Nvidia and model repositories including Hugging Face and ModelScope; NeMo Switchyard available on GitHub
1000000
—
—
open weights (public release)
Qwen3.8-2.4T-A95B
Alibaba Cloud Qwen team
3
50
—
Downloadable weights and configuration files for self-hosting; managed access also available through Qwen Cloud
262144
—
—
open weights
Qwen3.8-27B-CRACK-GGUF
Jinho Jang (dealign.ai)
3
50
—
Hugging Face repository; downloadable GGUF files and a 0.9 GB vision projector for local use
262000
—
—
Open weights; GGUF quantizations packaged for local inference via llama.cpp
GLM-5.3
Z.ai
5
83
—
Together AI serverless and dedicated infrastructure
1000000
1.400000
440.000000
API
Gemini 3.7 Flash
Google
4
67
—
Available via Google AI Studio, Android Studio, Google Antigravity, Gemini Enterprise Agent Platform, the Gemini Enterprise app, and Gemini Spark for Google AI Pro and Ultra subscribers in supported countries.
—
0.750000
3.750000
API and product availability
GLM-5.3
Zhipu (Z.ai)
0
0
—
—
—
—
—
—
GPT-5.6 Sol
OpenAI
2
33
—
Limited release to a small group of customers; wider access planned later
—
—
—
API
Ornith-1.5
Ornith AI
2
33
—
Hugging Face collection; local deployment and quantized mobile version described
—
—
—
open weights
Falcon 2
Murf AI
1
17
—
—
—
—
—
proprietary foundation voice model
V4-Flash-Vision-Exp
DeepSeek
2
33
—
DeepSeek API
—
—
—
experimental API model
deepseek-v4-flash-vision-exp
DeepSeek
4
67
—
Available through Chat Completions, Messages, and Responses requests
—
0.220000
0.660000
experimental API model
HiDream-O1-World
HiDream.ai
1
17
—
—
—
—
—
model launch
Thomson-1.0-Large
Thomson Reuters
1
17
—
—
—
—
—
proprietary model launch
Thomson
Thomson Reuters
3
50
Non-commercial license for the planned smaller open-weight version
Initially available in CoCounsel Legal’s Tabular Analysis feature; smaller version planned for Hugging Face
—
—
—
enterprise launch; open-weight version planned
GEN-1.5
Generalist AI
2
33
—
Available through direct partnerships on Generalist AI's own robot fleet; no public weights, API, or self-serve product
—
—
—
research release
Grok 4.6
xAI
5
83
—
Microsoft Foundry for Azure customers
500000
2.000000
6.000000
public preview
Granite 4.2
IBM
4
67
Apache 2.0
Hugging Face, Ollama, GitHub, and other platforms
512000
—
—
open weights
Qwen3.8-Flash-Next
Alibaba Qwen team
5
83
—
Weights on Hugging Face and ModelScope; production Qwen3.8-Flash through QwenCloud; API expected shortly
262144
0.160000
0.470000
architecture preview; open weights; production API
Qwen3.8-Flash
Alibaba
5
83
—
API access
262144
1.000000
3.000000
API
Qwen3.8-Flash
Alibaba Qwen
5
83
—
QwenCloud managed API
1000000
0.160000
0.470000
API
Gemini 3.5 Transcribe
Google
2
33
—
Public preview
—
—
—
API
GLM-5.3-Flash
Z.ai
6
100
MIT
Z.ai API; weights available on Hugging Face
1000000
0.150000
0.500000
API and open weights
GLM-5.3-Flash
Zhipu AI
6
100
MIT
Officially released; deployable on domestically produced AI chips
1000000
0.150000
0.500000
open-source
leiolai-1
Leiolai
5
83
—
Consumer app and developer API
11000000
0.010000
0.020000
consumer app and developer API
ABot-Recon
AMAP CV Lab
3
50
Apache 2.0 for source code; CC BY-NC 4.0 for model weights
Inference code, model checkpoint, and evaluation code released; weights limited to noncommercial research and education
—
—
—
open-source release
Sahara v2.5
Intron
1
17
—
—
—
—
—
model update
GigaPath-Flash and GigaTIME-Flash
Microsoft Research
3
50
Apache 2.0
Hugging Face
—
—
—
open weights; research release
S1
Skild AI
1
17
—
—
—
—
—
robot foundation model
Isaac 0.5
Perceptron AI
2
33
—
Model weights available through Hugging Face; fine-tuning and inference code available through GitHub
—
—
—
open weights
Orbis 1.0
Visko
2
33
—
public access
—
—
—
foundation model
Solaris
Runway
2
33
—
Early access via application form; launch partners sought
—
—
—
research
MAPL-EMIT
Google Research
2
33
—
Trained model and synthetic plume dataset on Kaggle; global plume database on Earth Engine; inference library on GitHub
—
—
—
trained model and research release
Claude Fable 5.1 and Claude Mythos 5.1
Anthropic
5
83
—
Claude Fable 5.1 is generally available through the Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, and Microsoft Foundry; Claude Mythos 5.1 is restricted to vetted U.S. organizations in Project Glasswing.
1000000
10.000000
50.000000
API and restricted preview
Muse Voice Transcribe
Meta Superintelligence Labs
2
33
—
Meta AI for Mac, Muse Code, and Meta Model API
—
—
—
App and API availability
Muse Spark 1.3
Meta
4
67
—
Broad rollout through Muse Code and Meta Model API for xhigh; max configuration limited to partner preview pending additional safety testing
—
1.250000
4.250000
API
Muse Spark 1.3
Meta AI Research
2
33
—
Rolling out in Muse Code and Meta Model API
—
—
—
API and integrated product rollout
Gemini 3.8 Flash; Gemini 3.8 Flash Cyber
Google
5
83
—
Gemini 3.8 Flash available through Google AI Studio and Android Studio; Gemini 3.8 Flash Cyber restricted to trusted governmental and critical-infrastructure entities through the Fairwind Program
1048576
0.750000
3.750000
launch
Gemini 3.8 Flash
Google
5
83
—
Gemini API, Google AI Studio, Android Studio, Google Antigravity, Gemini Enterprise, Gemini app, AI Mode in Google Search, and Gemini in Google Sheets
1000000
0.750000
3.750000
general availability
Gemini 3.8 Flash and Gemini 3.8 Flash Cyber
Google
5
83
—
Gemini 3.8 Flash is available in Gemini Enterprise and through Google AI Studio, Google Antigravity, Android Studio, and Stitch; Gemini 3.8 Flash Cyber is limited to trusted defenders through the Fairwind Program.
1000000
0.750000
3.750000
API, enterprise availability, and limited defender preview
K2 Horizon
Institute of Foundation Models
2
33
—
Hugging Face repositories
—
—
—
open weights
WeatherNext 3
Google DeepMind and Google Research
2
33
—
Global availability across Google Search, Gemini, Google Maps, Google Maps Platform Weather API, Google Earth Engine, BigQuery, and Google Cloud Storage
—
—
—
model launch
humain-m3
HUMAIN
3
50
MiniMax Community License (planned for open-weight release)
HUMAIN Node; limited-preview and research-preview tiers
—
—
—
research preview
GPT-6
OpenAI
2
33
—
Select cybersecurity customers initially; wider rollout to paying customers; unavailable to free-tier and cheapest-plan users
—
—
—
limited rollout
Lyria 3.5
Google
2
33
—
Gemini app, Google Flow Music, Google AI Studio, and Google Vids
—
—
—
API and app release
WeatherNext 3
Google DeepMind
2
33
—
Google Search, Gemini, Google Maps, Google Maps Platform Weather API, Google Earth Engine, BigQuery, and bulk download via Google Cloud Storage
—
—
—
general availability
Muse Voice Transcribe
Meta
2
33
—
Meta AI, Muse Code, and Meta Model API
—
—
—
API
Qwen-Drive 1.0
Alibaba
2
33
—
Free to the research community via Hugging Face, ModelScope, and GitHub
—
—
—
research release
Astra
OpenAI
2
33
—
Generally available
—
—
—
general availability
OUI-1
Thesys
3
50
Google Gemma terms
Local self-hosting; Thesys says it runs at FP8 on an Nvidia RTX 5090
—
—
—
open weights
N2.5 agent family
Nex-AGI
2
33
—
Mini and Max available; Pro weights marked coming soon
—
—
—
open weights
Suno v6
Suno
3
50
Trained on licensed content from Warner Music Group, BMG, and Believe; model license not specified
v6, v6-wild, and free tier v6-mini
—
—
—
model suite launch
V6
Suno
2
33
—
Available through three new subscription tiers
—
—
—
model launch
Granite PatchTST-FM-r2
IBM
4
67
Apache 2.0 or Linux Foundation OpenMDW 1.0
Downloadable from the Hugging Face Hub; available through granite-tsfm 0.3.9 or later
8192
—
—
open weights
GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst
OpenAI
4
67
—
Worldwide availability in ChatGPT, ChatGPT Work, Codex, and the API
—
8.000000
30.000000
API and ChatGPT release
Adaptive Instructed-Retriever
Databricks
1
17
—
—
—
—
—
model launch
DeepSeek-V4.1-Flash
DeepSeek
6
100
MIT
DeepSeek first-party API; US/API availability reported; also supported by Baseten and Ollama
1000000
0.300000
1.200000
open weights
NASA-IBM Lunar Foundation Model
NASA and IBM
3
50
Apache-2.0
Hugging Face weights; GitHub fine-tuning and inference code
—
—
—
open weights
Edge0 35B and Edge0 8B
Edge0
3
50
Apache 2.0
Released for macOS on Apple Silicon; no iOS build published
—
—
—
open weights and inference framework
Aether
JoyIn
2
33
—
Publicly announced; selected components planned for open-source release
—
—
—
robotics model announcement
GPT-Live-1
OpenAI
2
33
—
Available to developers
—
—
—
API
GPT-6 Astra
OpenAI
6
100
Proprietary
ChatGPT Plus, Pro, Business, and Enterprise; OpenAI API
1050000
10.000000
50.000000
API and product launch
Music v2.5
ElevenLabs
2
33
—
ElevenMusic app and API; free and Pro tiers
—
—
—
API and app
Iris-mini and Iris-pro
AllSpark
3
50
—
Model weights on Hugging Face; code and Iris Harness on GitHub
256000
—
—
open weights and open-source search-agent release
Boreal
Creatify Labs
2
33
—
Available now through the Creatify Labs API, Creatify Model Playground, AdFlow features, and fal platform
—
—
—
API
Gemini 3.8 Live
Google DeepMind
2
33
—
Available through the Gemini API, Google AI Studio, and Live API integrations
—
—
—
API
Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking
Google DeepMind
3
50
—
Gemini API, Gemini app, Google AI Studio, Google Cloud, Google Search Live, and selected Google Workspace surfaces; Live API documentation remains labeled preview
131072
—
—
API and platform release
Gemini 3.8 Live; Gemini 3.8 Live Extended Thinking
Google DeepMind
2
33
—
Available to developers through the Gemini API and Google AI Studio
—
—
—
API
TabPFN-3.5
Prior Labs
3
50
Non-commercial open-weight use; commercial production license or API required
Open weights available for research, evaluation, and Kaggle; production use through Prior Labs API or a commercial license
—
—
—
open weights
Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking
Google
4
67
—
Available in the Gemini Live API and Google AI Studio for production use
—
3.000000
12.000000
API
Pareto 26.9
unbiased.ai
5
83
—
OpenRouter and Cloudflare
262144
2.500000
7.500000
API
Koa
Salesforce
2
33
—
Piloting with Formula 1, Baxter Credit Union, and Xero; general availability planned for winter
—
—
—
preview
Grok Voice Transcribe 2.0
xAI
2
33
—
Available through REST and WebSocket Speech-to-Text API endpoints; selectable model, with planned default status
—
—
—
API
Bonsai 2 27B
PrismML
3
50
Apache 2.0
Weights available at launch; runs on Nvidia GPUs via CUDA and Apple devices via MLX
—
—
—
open weights
Ternary Bonsai 2 27B
PrismML
4
67
Apache 2.0
Released; local deployment via PrismML’s llama.cpp fork or MLX runtime
262000
—
—
open weights
Qwen3.8-Omni-Flash
Qwen
5
83
—
Available through Qwen Studio, Qwen Cloud, and the Qwen API
1000000
0.150000
0.470000
API
Qwen-Image-2.1
Alibaba Qwen AI team
3
50
Research license; commercial use requires a separate license from Qwen
Available on Hugging Face, GitHub, Model Scope, and via a Hugging Face demo
—
—
—
open weights
Alice AI Foundation
Yandex
3
50
Open license
Commercial and research use
—
—
—
open weights
Altar
Aikido
2
33
—
Available openly for local deployments
—
—
—
open weights
RetroChimera
Microsoft Research
3
50
MIT
GitHub and Microsoft Foundry
—
—
—
open weights
Qwen-Image-2.1
Alibaba Qwen
3
50
Qwen Research License
Research and evaluation; commercial deployment requires a separate Qwen license
—
—
—
open weights
Grok 4.7
SpaceXAI
5
83
—
Available in the Grok API, Cursor, Grok Build, third-party coding tools, model routers, and cloud platforms; gradual GitHub Copilot rollout
500000
2.000000
6.000000
API and product availability
Grok 4.7
SpaceX
4
67
—
Available from $2 per million input tokens and $6 per million output tokens
—
2.000000
6.000000
API
Xing4.0-29B-A4B
China Telecom AI
3
50
—
GitHub and Hugging Face repositories
256000
—
—
open weights
MiMo-V2.6-Pro
Xiaomi
6
100
MIT
Weights downloadable from Xiaomi's Hugging Face repository; API access listed through OpenRouter and other resellers.
1000000
0.435000
0.870000
open weights
GPT-6 Sol and GPT-6 Luna
OpenAI
3
50
—
API; ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users; Luna also available in the ChatGPT desktop app for Free and Go subscribers
1050000
—
—
API and product rollout
Claude Opus 5.5
Anthropic
5
83
—
Anthropic API, AWS, Google Cloud, and Microsoft’s ecosystem
1000000
4.000000
20.000000
API
Nemotron 3 Diarization
NVIDIA
3
50
OpenMDW License 1.1
Available on Hugging Face
—
—
—
open weights
FLUX 3 Action
Black Forest Labs
2
33
—
Previously accessible through selected research and commercial partners; public weights were not yet available according to the article
—
—
—
Robotics model; broader open-weight release announced as forthcoming
Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS
Google
2
33
—
Rolling out through the Gemini API and Google AI Studio; enterprise API access via Gemini Enterprise is listed as coming soon.
—
—
—
API
Opus 5.5
Anthropic
2
33
—
Launched for Claude; special permission required for cyber security and biology researchers using it in their work
—
—
—
model launch
Measurement technique
How to read this report
01The score measures whether six practical launch fields are disclosed: license, availability, context window, input price, output price, and release type.
02Every field receives equal weight; missing is distinct from zero and receives no credit.
03This is a disclosure-completeness score, not a safety, quality, or trustworthiness rating.
Sources
Evidence
26 publishers supporting 41 records. Expand a publisher to inspect its cited pages.
the-decoder.comthe-decoder.com6 sources / 6 records