Home › By model
By model and version, when the post names it.
Most posts say “ChatGPT” or “Claude” and stop there. When a post names the model or version (GPT-4o, o3, Claude 3.5 Sonnet, Gemini 2.x, Lexis+ AI Protégé…), we file it here too, with its date, because the tools change under the same name. 632 model mentions across 19 models and versions as of Oct 26, 2026.
Dated report: Frontier models for legal work, Claude vs ChatGPT vs Gemini, our record and public benchmarks kept apart (PDF).
Public benchmarks, dated and summarised: Legal AI benchmarks (Vals VLAIR, Stanford hallucination study, LegalBench), other publishers’ figures kept apart from ours (PDF).
ChatGPT
Net only from posts whose main subject is ChatGPT, and only from 5 such posts.
| Model or version | Family | Posts naming it | Share of family’s model posts | About the family tool | Dislike per 100 90% band | Net 90% band | 2023 | 2024 | 2025 | 2026* | First – last |
|---|---|---|---|---|---|---|---|---|---|---|---|
| GPT-4 |  ChatGPT | 95 | 41% | 65 | 28 19–37 | +12 −4 to +28 | 49 | 27 | 12 | 7 | Jan 16, 2023 – Sep 30, 2026 |
| GPT-5 | ChatGPT | 57 | 25% | 23 | 30 17–47 | +13 −16 to +39 | 25 | 32 | May 25, 2025 – Sep 29, 2026 | ||
| GPT-4o | ChatGPT | 40 | 17% | 23 | 17 7–32 | +43 +13 to +65 | 14 | 18 | 8 | May 15, 2024 – Sep 30, 2026 | |
| o3 | ChatGPT | 20 | 9% | 11 | 9 2–31 | +82 +37 to +95 | 20 | Apr 17, 2025 – Sep 13, 2025 | |||
| o1 | ChatGPT | 10 | 4% | 7 | 43 18–72 | 0 −47 to +47 | 2 | 7 | 1 | Oct 20, 2024 – Feb 18, 2026 | |
| GPT-3.5 | ChatGPT | 7 | 3% | 7 | 29 9–58 | +14 −35 to +55 | 6 | 1 | Jan 9, 2023 – Jun 6, 2026 |
CoCounsel
Net only from posts whose main subject is CoCounsel, and only from 5 such posts.
| Model or version | Family | Posts naming it | Share of family’s model posts | About the family tool | Dislike per 100 90% band | Net 90% band | 2023 | 2024 | 2025 | 2026* | First – last |
|---|---|---|---|---|---|---|---|---|---|---|---|
| CoCounsel (Casetext era) | CoCounsel | 117 | 100% | 75 | 19 12–27 | +33 +17 to +47 | 25 | 38 | 41 | 13 | Mar 11, 2023 – Sep 18, 2026 |
Lexis+ AI
Net only from posts whose main subject is Lexis+ AI, and only from 5 such posts.
| Model or version | Family | Posts naming it | Share of family’s model posts | About the family tool | Dislike per 100 90% band | Net 90% band | 2023 | 2024 | 2025 | 2026* | First – last |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Lexis+ AI Protégé | Lexis+ AI | 94 | 100% | 65 | 35 26–46 | +6 −11 to +23 | 30 | 64 | Feb 2, 2025 – Oct 17, 2026 |
Gemini
Net only from posts whose main subject is Gemini, and only from 5 such posts.
| Model or version | Family | Posts naming it | Share of family’s model posts | About the family tool | Dislike per 100 90% band | Net 90% band | 2023 | 2024 | 2025 | 2026* | First – last |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Gemini 2.x | Gemini | 29 | 32% | 14 | 7 1–24 | +57 +22 to +76 | 25 | 4 | Feb 14, 2025 – Aug 5, 2026 | ||
| Bard | Gemini | 27 | 30% | 11 | 64 39–83 | −36 −69 to +11 | 17 | 6 | 4 | May 12, 2023 – May 14, 2025 | |
| Gemini 3 | Gemini | 22 | 24% | 10 | 0 0–17 | +80 +41 to +91 | 9 | 13 | Jul 30, 2025 – Sep 17, 2026 | ||
| Gemini Advanced / Pro | Gemini | 12 | 13% | 7 | 29 9–60 | +29 −27 to +68 | 4 | 6 | 2 | Feb 26, 2024 – Aug 24, 2026 |
Claude
Net only from posts whose main subject is Claude, and only from 5 such posts.
| Model or version | Family | Posts naming it | Share of family’s model posts | About the family tool | Dislike per 100 90% band | Net 90% band | 2023 | 2024 | 2025 | 2026* | First – last |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Claude Opus 4.x | Claude | 27 | 43% | 15 | 13 4–33 | +47 +12 to +70 | 4 | 23 | Jul 31, 2025 – Oct 6, 2026 | ||
| Claude 3.5 Sonnet | Claude | 12 | 19% | 10 | 0 0–17 | +100 +61 to +99 | 7 | 4 | 1 | Jul 19, 2024 – Jun 7, 2026 | |
| Claude Opus (unversioned) | Claude | 11 | 17% | 7 | 0 0–23 | +86 +34 to +94 | 1 | 4 | 6 | Apr 29, 2024 – Sep 3, 2026 | |
| Claude Sonnet 4.x | Claude | 10 | 16% | 5 | 20 4–56 | +20 −33 to +59 | 4 | 6 | Jul 27, 2025 – Aug 5, 2026 | ||
| Claude 3.7 Sonnet | Claude | 3 | 5% | 2 | too few | too few | 3 | Mar 14, 2025 – May 4, 2025 |
Westlaw AI
Net only from posts whose main subject is Westlaw AI, and only from 5 such posts.
| Model or version | Family | Posts naming it | Share of family’s model posts | About the family tool | Dislike per 100 90% band | Net 90% band | 2023 | 2024 | 2025 | 2026* | First – last |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Westlaw Precision | Westlaw AI | 32 | 82% | 18 | 39 22–59 | 0 −32 to +33 | 13 | 14 | 5 | Feb 19, 2024 – May 8, 2026 | |
| Westlaw Edge | Westlaw AI | 7 | 18% | 1 | too few | too few | 1 | 2 | 1 | 3 | May 13, 2023 – Apr 21, 2026 |
* current year to date. “Posts naming it” counts every voice whose text names the model, whatever tool the post is mainly about (a comparison post counts for both). The net uses only posts whose main tool is that family, so it is that family’s reception while people were naming that version. Small n: read the band, not the dot. More posts naming a model does not mean it is worse: compare the share of posts and the dislike per 100 posts about that family, never the raw count of complaints. The evolution of each model quarter by quarter is on Evolution.
Which model the tool says it uses
Only what the vendor or the model provider says on its own pages, read on Oct 22, 2026. “Not disclosed” means we found no official statement.
| Tool | Model | Status | Period | What the official page says |
|---|---|---|---|---|
| Harvey | Anthropic Claude Fable 5.1 | Offers | as of 2026-09-01 | Harvey announced Anthropic's Claude Fable 5.1 is available in Harvey. Source Fable 5.1, Now Live in Harvey, Sep 1, 2026 |
| Harvey | Claude Opus 4.6, GPT-5.2, Sonnet 4.6, Gemini 3 Flash (Anthropic, OpenAI, Google) | Uses | as of 2026-03-10 | Harvey describes a multi-model setup with foundation models from Anthropic, Google DeepMind and OpenAI, naming Opus 4.6, GPT-5.2, Sonnet 4.6 and Gemini 3 Flash. Source Why Harvey is Multi-Model by Design, Mar 10, 2026 |
| Harvey | Anthropic and Google models added alongside OpenAI | Uses | since 2025-05 | Harvey announced adding Anthropic and Google models, via AWS Bedrock and Google Vertex, to supplement its existing OpenAI models. Source Expanding Harvey's Model Offerings, May 13, 2025 |
| Harvey | OpenAI o1 | Used | 2024-09 | Harvey said it would build legal agents and workflows with OpenAI's o1 reasoning series. Source Harvey is building legal agents and workflows with OpenAI o1, Sep 12, 2024 |
| CoCounsel | Anthropic Claude Agent SDK (next-gen CoCounsel Legal) | Uses | as of 2026-08-20 | Thomson Reuters says the next-generation CoCounsel Legal is built on Anthropic's Claude Agent SDK; Tabular Analysis uses a TR proprietary legal LLM. Source Thomson Reuters Launches Next Generation of CoCounsel Legal, Aug 20, 2026 |
| CoCounsel | Claude (Anthropic), GPT (OpenAI), Gemini (Google) plus proprietary | Uses | as of 2026-02-24 | TR states it works with frontier models from Anthropic, OpenAI and Google combined with its own proprietary AI. Source One Million Professionals Turn to CoCounsel as Thomson Reuters Scales AI for Regulated Industries, Feb 24, 2026 |
| CoCounsel | OpenAI o1-mini (custom), GPT-4, Claude 3.5 (evaluated), Gemini 1.5 Pro | Used | 2023-2024 | TR said CoCounsel was tested with a custom OpenAI o1-mini model, launched on GPT-4, was evaluating Claude 3.5 models, and integrated Gemini 1.5 Pro. Source Thomson Reuters CoCounsel tests custom LLM from OpenAI, broadening its multi-model product strategy, Nov 25, 2024 |
| CoCounsel | Google Gemini 1.5 Pro (testing) | Used | 2024-08 | CoCounsel 2.0 release said TR was testing adding Gemini 1.5 Pro to its production models suite, alongside OpenAI. Source Thomson Reuters unveils CoCounsel 2.0, Aug 12, 2024 |
| Lexis+ AI | Protégé General AI: Claude Sonnet 4.5, GPT-5.1, plus earlier Claude Sonnet 4, GPT-4o, GPT-5, o3 | Offers | as of 2025-12-10 | LexisNexis says Protégé General AI users can choose among Anthropic and OpenAI models or use the automatic Best Fit default. Source LexisNexis Unveils Next-Generation Protégé General AI, Dec 10, 2025 |
| Lexis+ AI | Anthropic Claude legal plugin suite (12 plugins) | Uses | as of 2026-05-13 | LexisNexis announced integrating Anthropic's Claude legal plugin suite into Lexis+ with Protégé. Source LexisNexis Expands Lexis+ with Protégé by Integrating Anthropic's Claude Legal Plugin Suite, May 13, 2026 |
| Lexis+ AI | Protégé Legal AI: Claude 3.7, Claude 4, GPT-4o, GPT-4.1, fine-tuned GPT-4o-mini, fine-tuned Mistral | Uses | as of 2025-11-25 | LexisNexis lists multiple legal-tuned models behind Protégé Legal AI (Malaysia press-room reprint). Source LexisNexis Introduces Protégé General AI and Expands Agentic AI Leadership, Nov 25, 2025 |
| Westlaw AI | not named (a large language model) | Not disclosed | as of 2023-11-15 | TR's launch blog for Westlaw AI-Assisted Research says it uses an LLM but does not name the model. Later official CoCounsel pages name Claude, GPT, Gemini generally, not Westlaw specifically. Source Legal research meets generative AI, Nov 15, 2023 |
| Legora | Anthropic Claude (Sonnet 4 evaluated) | Uses | undated Anthropic case study (page footer 2026) | Anthropic's customer page says Legora uses Claude across its platform; it also reports Claude Sonnet 4 results on Legora's evaluation set. No OpenAI/other statement found in official sources read. Source Legora case study | Claude, source page undated, read Oct 22, 2026 |
| Spellbook | GPT-5 (OpenAI per vendor page; provider not stated on page) | Uses | as of 2026 (page undated) | Spellbook's own pages advertise GPT-5 in Word; the pages do not name the provider and no Claude statement was found. Source Spellbook: Legal AI in Word, source page undated, read Oct 22, 2026 |
| Spellbook | GPT-5 | Uses | as of 2026 (page undated) | Spellbook has a page announcing GPT-5 for lawyers. Source GPT-5 for Lawyers is now available | Spellbook, source page undated, read Oct 22, 2026 |
| Clio | Not read | Not disclosed | as of 2026-10-22 | No model statement read: Clio's own pages refuse automated reads (HTTP 403), so we state nothing about its models. Source Clio press page (403), source page undated, read Oct 22, 2026 |
| vLex Vincent | OpenAI GPT, Google Gemini, Anthropic Claude | Uses | as of 2026 (page undated) | vLex support page says Vincent draws on models from OpenAI, Google and Anthropic. Source Understanding the AI Models Used by Vincent, source page undated, read Oct 22, 2026 |
| Luminance | Proprietary Luna Crescent plus multi-model; tests GPT 6 Sol/6.1 Sol and Claude Sonnet 5.5/Opus 5.5 | Uses | as of 2026-10-05 | Luminance says it evaluated OpenAI and Anthropic models for Contract Intelligence; final validation pending before production. Its tech page names no providers. Source How Luminance selects AI models for Contract Intelligence, Oct 5, 2026 |
| Paxton | not named | Not disclosed | as of 2026-10-22 | Paxton names no model provider on its public home page. Source Paxton, source page undated, read Oct 22, 2026 |
| Copilot | OpenAI latest models plus Anthropic Claude Sonnet 4 and Opus 4.1 | Offers | as of 2025-09-24 | Microsoft says Copilot stays on OpenAI's latest models and adds Anthropic Claude options; Researcher can use OpenAI or Claude Opus 4.1. Source Expanding model choice in Microsoft 365 Copilot, Sep 24, 2025 |
| Copilot | OpenAI GPT-5 | Uses | since 2025-08-07 | Microsoft announced GPT-5 available in Microsoft 365 Copilot. Source Available today: GPT-5 in Microsoft 365 Copilot, Aug 7, 2025 |
| Perplexity | Anthropic Claude, OpenAI GPT, Google Gemini (Agent API model list) | Offers | as of 2026-05-21 | Perplexity's API docs (not the consumer help center, which returned 403) list Anthropic, OpenAI and Google models; this is the Agent API, not necessarily the app's Pro menu. Source Agent API Models - Perplexity, May 21, 2026 |
Use these numbers: the counts, rates and bands on this page are free to reuse under CC BY 4.0. Credit “aiforlawyers.app (AI Vortex)” with the date and a link to the page. Quotes belong to the people who wrote them: quote them with the link to the original post. Machine-readable: index.json. How to cite.