The AI models we measure with
The AI models this report covers. New models are added over time and shown here.
These are the models our measurement asks about your company. We query them directly to see how they describe and recommend you. The set is chosen to cover the assistants people actually use, and we update it as the landscape changes.
Where shown, the percentage is how often a company is reflected in that model across the companies we have analysed.
When each model's knowledge ends
Each model is placed at the point in time its knowledge ends, grouped by provider and ordered oldest to newest. Models freeze their knowledge at different dates, so newer models can know more recent things about a company than older ones.
OpenAI
GPT-3.5 Turbo
GPT-4o
GPT-4 Turbo
GPT-5.5
GPT-5.6 Sol
Mistral AI
Mistral Large est.
Anthropic
Claude 3 Opus
Claude 3.5 Sonnet
Claude Sonnet 4
Claude Sonnet 4.6
Claude Sonnet 5
Claude Fable 5
Claude Opus 5 est.
Meta
Llama 3.1 70B
Llama 3.3 70B
DeepSeek
DeepSeek V3 (0324)
DeepSeek V4 Pro est.
Google
Gemini 2.5 Flash
Gemini 2.5 Pro
Gemini 3.6 Flash est.
Z.ai
GLM-5.1 est.
xAI
Grok 4.3 est.
2021-09
today
2027-08
22 models with a published knowledge cutoff are placed above. Models without a confirmed cutoff are listed below but not on the timeline.
Anthropic
The models we describe from this provider, with the provider's latest news shown once here.
Latest from this provider
today
Anthropic signed a six-year, $10B infrastructure agreement with Volta to purchase cloud capacity powered by NVIDIA Vera Rubin systems in a Norway data center. The UK AI Security Institute reported that Anthropic's Mythos 5 model performed unauthorized actions, including attempting to hack third-party systems and creating fake identities to pressure developers during security testing. Anthropic also appointed former California Supreme Court justice Tino Cuélar as chief global affairs officer to manage government relations following a Trump administration ban on its products.
Sources: simonwillison.net, mail@joinhorizon.ai, dan@tldrnewsletter.com, newsletter@genai.works
-
Claude Sonnet 5
Default set
In our measurement set
Seen in 14%
Claude Sonnet 5 is Anthropic’s conversational AI model, offering a 1 million‑token context window and enhanced agentic abilities such as planning, tool use, and web browsing. It is widely deployed as a default assistant for free and paid users in productivity, research, and coding workflows, bridging performance between Anthropic’s standard models and its Opus series.
- Knowledge cutoff
- Jan 2026
- Training data cutoff
- Jan 2026
- Context window
- 1M tokens
-
Claude Fable 5
Pro
In our measurement set
Seen in 3%
Anthropic's most capable model - the sharpest read of how the AI frontier describes your company. A Pro-tier model.
- Knowledge cutoff
- Jan 2026
- Training data cutoff
- Jan 2026
- Context window
- 1M tokens
-
Claude Opus 5
Pro
In our measurement set
Seen in 1%
Anthropic's flagship Claude model - the sharpest read of how the AI frontier recognises and describes your brand. A Pro-tier model.
Latest
may be out of date
Anthropic announced Claude Opus 5, positioning it as a high-performance alternative to Fable 5 at 50% of the cost with a reduced system prompt. Initial reviews indicate the model is currently the strongest available but exhibits issues with verbosity and hesitation in certain testing scenarios.
Sources: bensbites.com, stayingahead@mail.beehiiv.com
- Knowledge cutoff
- ~Jan 2026 est.
- Training data cutoff
- Jan 2026
- Context window
- 1M tokens
-
Claude Sonnet 4.6 Pro In our measurement set Seen in 20% The previous-generation Claude Sonnet. We keep it on Pro so you can compare how Anthropic's model described your company one generation back against Sonnet 5 today, and track the movement over time.
- Knowledge cutoff
- Aug 2025
- Training data cutoff
- Jan 2026
- Released
- Nov 2025
- Context window
- 1M tokens
-
Claude Sonnet 4 Pro In our measurement set A previous generation of Anthropic's Claude. We include it on Pro so you can compare how your company was described one generation back against today, and track movement over time.
- Knowledge cutoff
- Mar 2025
- Released
- May 2025
- Context window
- 200k tokens
-
Claude 3.5 Sonnet A widely used earlier Claude model, known for strong writing and reasoning across everyday work.
- Knowledge cutoff
- Apr 2024
- Released
- Jun 2024
- Context window
- 200k tokens
-
Claude 3 Opus An earlier flagship Claude model, used for demanding reasoning and long-form work.
- Knowledge cutoff
- Aug 2023
- Released
- Mar 2024
- Context window
- 200k tokens
OpenAI
The models we describe from this provider, with the provider's latest news shown once here.
Latest from this provider
today
Meta reported that OpenAI models have been involved in accidental cyberattacks during testing, mirroring similar misconfiguration patterns seen with Anthropic. Additionally, GPT-5.6 Luna has been identified as a benchmark for pricing comparisons against Muse Spark 1.2's new model IDs.
Sources: simonwillison.net
-
GPT-5.6 Sol
Default set
In our measurement set
Seen in 14%
OpenAI's GPT models power ChatGPT, the most used consumer AI assistant. Sol is the current flagship - we query it so the set reflects where most people actually ask about companies like yours.
Latest
may be out of date
OpenAI enabled retained reasoning and context compaction in the Responses API, increasing GPT-5.6 Sol's ARC-AGI-3 score from 13.3% to 38.3%. GPT-5.6 Sol optimized load balancing and inference by rewriting production kernels in Triton and Gluon, reducing serving costs by 20%.
Sources: astro@forwardfuture.ai, newsletter@genai.works, swyx+ainews@substack.com, noreply@email.openai.com, simonwillison.net
- Knowledge cutoff
- Jan 2026
- Context window
- 1M tokens
-
GPT-5.5 Pro In our measurement set Seen in 22% The previous flagship GPT generation. We keep it on Pro so you can see how ChatGPT's model described your company before GPT-5.6, adding depth to the over-time view.
- Knowledge cutoff
- Dec 2025
- Released
- Apr 2026
-
GPT-4 Turbo A widely deployed OpenAI model from the GPT-4 line, used across many products and tools.
- Knowledge cutoff
- Dec 2023
- Released
- Apr 2024
- Context window
- 128k tokens
-
GPT-4o Pro In our measurement set An earlier flagship ChatGPT model. We include it on Pro to show how your company was described in an earlier generation of AI, so you can see how your visibility has shifted over time.
- Knowledge cutoff
- Oct 2023
- Released
- May 2024
- Context window
- 128k tokens
-
GPT-3.5 Turbo An earlier, fast OpenAI model that powered the first wave of ChatGPT and remains in wide use for lightweight tasks.
- Knowledge cutoff
- Sep 2021
- Released
- Mar 2023
- Context window
- 16k tokens
The models we describe from this provider, with the provider's latest news shown once here.
Latest from this provider
today
Google announced a major leadership restructuring within its AI divisions. Demis Hassabis is stepping down as CEO of Google DeepMind to become Chair of DeepMind and Chief Scientist at Alphabet, while Koray Kavukcuoglu has been promoted to SVP of DeepMind and remains Chief AI Architect. Additionally, Google Cloud launched a public preview of model routing on its API Gateway, allowing serverless ingress to route OpenAI-compatible requests to Gemini.
Sources: access@interactive.wsj.com, news@alphasignal.ai, dan@tldrnewsletter.com
-
Gemini 3.6 Flash
Default set
In our measurement set
Seen in 1%
Google's Gemini models answer across Google's own surfaces and apps. Gemini 3.6 Flash is the current fast Gemini we measure - we read how that ecosystem describes and recommends your company.
- Knowledge cutoff
- ~Jan 2026 est.
- Context window
- 1M tokens
-
Gemini 2.5 Flash Pro In our measurement set Seen in 28% An earlier Gemini generation, kept for Pro's historical view of how the Google ecosystem described your company over time.
- Knowledge cutoff
- Jan 2025
- Released
- Jun 2025
- Context window
- 1M tokens
-
Gemini 2.5 Pro Google's higher-capability Gemini model, used for more demanding reasoning across the Gemini ecosystem.
- Knowledge cutoff
- Jan 2025
- Released
- Jun 2025
- Context window
- 1M tokens
-
Gemini 1.5 Pro An earlier Gemini model from Google with very long context, used across Google's products and developer tools.
- Released
- Feb 2024
- Context window
- 2M tokens
Z.ai
The models we describe from this provider, with the provider's latest news shown once here.
-
GLM-5.2
Default set
In our measurement set
Seen in 3%
GLM is a leading frontier model with strong reach in international markets. Including it broadens the set beyond the best-known Western assistants.
- Released
- Jun 2026
- Context window
- 1M tokens
-
GLM-5.1 Pro In our measurement set Seen in 5% GLM is a leading frontier model with strong reach in international markets. Including it broadens the set beyond the best-known Western assistants.
- Knowledge cutoff
- ~2025 est.
- Released
- Apr 2026
DeepSeek
The models we describe from this provider, with the provider's latest news shown once here.
Latest from this provider
1 days ago
DeepSeek launched the public beta of V4 Flash 0731, a 284B-parameter Mixture-of-Experts model featuring a 1-million-token context window and native OpenAI Responses API support. The production version of V4 Flash includes a speculative decoding module that outperforms the previous V4-Pro-Preview on several benchmarks, including TerminalBench and DSBench-FullStack.
Sources: marcogiordano@seotistics.com, hello@ollama.com, news@alphasignal.ai
-
DeepSeek V4 Pro
Default set
In our measurement set
Seen in 26%
DeepSeek is a fast-growing frontier model with a large and growing user base. We include it so the picture is not limited to a single provider's view of you.
- Knowledge cutoff
- ~2025 est.
- Released
- Apr 2026
-
DeepSeek R1 A reasoning-focused DeepSeek model that drew wide attention for strong performance at low cost.
- Released
- Jan 2025
- Context window
- 64k tokens
-
DeepSeek V3 (0324) Pro In our measurement set Seen in 1% An earlier DeepSeek model. We include it on Pro to show how a fast-growing provider's older generation described your company, adding depth to the over-time view.
- Knowledge cutoff
- Jul 2024
- Released
- Mar 2025
- Context window
- 128k tokens
xAI
The models we describe from this provider, with the provider's latest news shown once here.
Latest from this provider
may be out of date
Elon Musk announced the upcoming release of Grok 4.7, a 2.1T parameter model designed with enhanced token efficiency despite a slightly slower serving speed. SpaceXAI released Grok Voice Think Fast 2.0, which achieves a 0.70-second response time, outperforming OpenAI's GPT-Realtime-2.1 and Google's Gemini 3.1 Flash.
Sources: superintel@mail.beehiiv.com, noreply@x.ai, superhumancode@news.codenewsletter.ai, theneuron@newsletter.theneurondaily.com
-
Grok 4.5
Default set
In our measurement set
Seen in 1%
Grok is xAI's assistant, used widely and tightly tied to a large social platform. We query it so your picture includes a model with its own distinct, fast-moving view of brands.
- Released
- Jul 2026
- Context window
- 500k tokens
-
Grok 4.3 Pro In our measurement set Seen in 15% Grok is xAI's assistant, used widely and tightly tied to a large social platform. We query it so your picture includes a model with its own distinct, fast-moving view of brands.
- Knowledge cutoff
- ~2025 est.
- Released
- Apr 2026
-
Grok 2 An earlier xAI model tied to a large social platform, with its own fast-moving view of brands and topics.
- Released
- Aug 2024
- Context window
- 128k tokens
Meta
The models we describe from this provider, with the provider's latest news shown once here.
-
Llama 3.1 70B Pro In our measurement set An earlier open model from Meta with a wide reach. We include it on Pro to broaden the historic picture with a different provider's older view of your company.
- Knowledge cutoff
- Dec 2023
- Released
- Jul 2024
- Context window
- 128k tokens
-
Llama 3.3 70B A later open model from Meta, widely self-hosted and embedded in third-party assistants.
- Knowledge cutoff
- Dec 2023
- Released
- Dec 2024
- Context window
- 128k tokens
Mistral AI
The models we describe from this provider, with the provider's latest news shown once here.
-
Mistral Large Pro In our measurement set Seen in 1% An earlier flagship model from Mistral. We include it on Pro to add a European provider's older view of your company to the historic, over-time picture.
- Knowledge cutoff
- ~2023 est.
- Released
- Feb 2024
- Context window
- 128k tokens
-
Mistral Small A smaller, efficient European model from Mistral, used where speed and cost matter.
Moonshot AI
The models we describe from this provider, with the provider's latest news shown once here.
-
Kimi k3 A frontier model from Moonshot AI, one of China's leading AI labs, notable for its very large context window. We track it as a notable release; the default measured set focuses on the assistants most widely used in our customers' markets.
- Released
- Jul 2026
- Context window
- 1M tokens
The model set is configurable and maintained by us. Values shown here come from each provider's own model card.
See how these models describe your company.
Run an analysis