Verify any claim · lenz.io
“GPT-4 had a context window of 128,000 tokens at its release in March 2023.”
The conclusion
The March 2023 release did not provide a 128,000-token context window. OpenAI documented an 8,192-token window for GPT-4 and limited access to a 32,768-token variant; the 128K window was announced for GPT-4 Turbo in November 2023. The claim conflates the original model with that later release.
Caveats
- The claim conflates the original GPT-4 with the later GPT-4 Turbo model.
- The 128K context window was announced in November 2023, not March 2023.
- General references to long-context research do not establish a 128K launch capacity.
Fact-check inside the tools you already use
Connect Lenz to ChatGPT, Claude, or WhatsApp and check a claim mid-conversation.
Or create a free account to bookmark this verification and run your own checks.
Sources
Sources used in the analysis
Today all existing API developers with a history of successful payments can access the GPT‑4 API with 8K context.
Today all existing API developers with a history of successful payments can access the GPT-4 API with 8K context.
GPT-4 is an older version of a high-intelligence GPT model, usable in Chat Completions. 8,192 context window
gpt-4 has a context length of 8,192 tokens. We are also providing limited access to our 32,768–context (about 50 pages of text) version, gpt-4-32k, which will also be updated automatically over time (current version gpt-4-32k-0314, also supported until June 14).
We released the first version of GPT‑4 in March and made GPT‑4 generally available to all developers in July. Today we’re launching a preview of the next generation of this model, GPT‑4 Turbo(opens in a new window). … GPT‑4 Turbo is more capable and has knowledge of world events up to April 2023. It has a 128k context window so it can fit the equivalent of more than 300 pages of text in a single prompt.
gpt-4 has a context length of 8,192 tokens. We are also providing limited access to our 32,768–context (about 50 pages of text) version, gpt-4-32k, which will also be updated automatically over time (current version gpt-4-32k-0314, also supported until June 14).
gpt-4 has a context length of 8,192 tokens. We are also providing limited access to our 32,768–context (about 50 pages of text) version, gpt-4-32k, which will also be updated automatically over time (current version gpt-4-32k-0314, also supported until June 14).
Despite its capabilities, GPT-4 has similar limitations to earlier GPT models [1, 37, 38]: it is not fully reliable (e.g. can suffer from “hallucinations”), has a limited context window, and does not learn from experience.
OpenAI first introduced the 32K model when it unveiled GPT-4 in March, but limited access first to select users and then to the API, likely for cost reasons.
However, GPT-4 has significant limitations for cybersecurity operations due to its “hallucination” tendency and limited context window.
OpenAI has released a powerful new image- and text-understanding AI model, GPT-4, that the company calls “the latest milestone in its effort in scaling up deep learning.” … Pricing is $0.03 per 1,000 “prompt” tokens (about 750 words) and $0.06 per 1,000 “completion” tokens (again, about 750 words).
GPT-4 was released on March 14, 2023. The default version had a 8K context window, with a special version in the API supporting up to 32K tokens. … In November 2023, OpenAI announced the GPT-4 Turbo and GPT-4 Turbo with Vision model, which features a 128K context window and significantly cheaper pricing.
OpenAI has recently introduced the GPT-4 Turbo model with a 128K context window. A 4x boost to the previous maximum of 32K in GPT-4. … November 6, 2023 : GPT-4 Turbo - 128K … March 14, 2023 : GPT4 - 8K and 32K
Long context Core contributors 11 Gabriel Goh Long context co-lead Łukasz Kaiser Long context lead Ben Wang Attention architecture lead Clemens Winter Long context co-lead Long context research 11 Mo Bavarian, Gabriel Goh, Heewoo Jun, Łukasz Kaiser, Chak Ming Li, Ben Wang, Clemens Winter Long context kernels 11 Phil Tillet
Despite its capabilities, GPT-4 has similar limitations to earlier GPT models [1, 31, 32]: it is not fully reliable (e.g. can suffer from “hallucinations”), has a limited context window, and does not learn ∗Please cite this work as “OpenAI (2023)".
The models page in the openAI documentation says chat optimized GPT-4 has a context length of 8192. … They are also planning to come out with a GPT-4-32k, but that’s not available right now.
The number of total tokens (input + output) allowed by OpenAI depends on the model you use: 4k for `text-davinci-003`, 8k for `gpt-4`, etc: … 1. Context length (where GPT-3.5 = 4K and GPT-4 = 8K/32K) refers to ALL TOKENS (words, spaces, sentences, whatever) used in ALL REQUESTS until the model “forgets” the entire conversation, INCLUDING the INPUT (e.g user questions) and OUTPUT (e.g model answers) ?
The invitation email tells me that I can use GPT-4 with 8K context.
Our standard GPT-4 model offers 8,000 tokens for the context. We also offer an extended 32,000 token context-length model, which we are rolling out separately to the 8k model.
Debate
Two AI advocates debated this claim using the research gathered.
Argument for
OpenAI describes GPT-4 Turbo as the next generation of GPT-4 and assigns it a 128K context window, establishing 128K as a GPT-4 capability within the model line (Source 5, openai.com). The GPT-4 technical materials' emphasis on long-context work supports treating the 128K implementation as part of GPT-4's released architecture rather than a distinct non-GPT-4 system (Sources 8 and 14).
The Proponent's own cited Source 5 explicitly dates the 128k context window to a November 2023 DevDay 'preview' of GPT-4 Turbo, a distinct model release eight months after the March 2023 launch, directly contradicting the claim's specified release date. Source 14's mention of a 'long context' research team is not evidence of a 128k window at launch—this is a non sequitur, since Sources 4, 6, 7, and the Technical Report itself (verified against the original March 14 announcement) unambiguously state GPT-4 shipped with 8,192 and 32,768-token windows, not 128,000.
Argument against
The claim is decisively refuted by OpenAI's own original March 14, 2023 release announcement (Sources 4, 6, 7) and the GPT-4 Technical Report, which explicitly state GPT-4 launched with an 8,192-token context length, with a limited 32,768-token variant—not 128,000 tokens. Every corroborating source, including Wikipedia (Source 12) and dev.to's timeline (Source 13), confirms the 128K context window belongs exclusively to GPT-4 Turbo, announced at DevDay in November 2023 (Source 5), meaning the claim conflates two distinct models separated by eight months and is unambiguously false.
The Opponent commits an equivocation by treating GPT-4 Turbo as categorically outside GPT-4 even though OpenAI expressly identifies it as the “next generation” of GPT-4 and attributes the 128K window to that GPT-4 implementation (Source 5, openai.com). The Opponent also overstates the Technical Report: the cited passage merely notes a “limited context window,” whereas OpenAI's long-context research and GPT-4 Turbo announcement support viewing 128K as an evolved GPT-4 capability rather than evidence of a wholly separate system (Sources 8, 14, and 5).
Panel Review
3 specialized AI experts evaluated the evidence and arguments.
Reviewer 1 — The Logic Examiner
The evidence clearly shows that GPT-4 was released in March 2023 with an 8K context window and a limited 32K version (Sources 4, 6, 7, 12, 13). The 128K context window was introduced with GPT-4 Turbo in November 2023 (Sources 5, 12, 13), making the claim's assertion about the March 2023 release date logically false.
Reviewer 2 — The Source Auditor
OpenAI's own original release materials from March 14, 2023 (Sources 4, 6, 7) explicitly state GPT-4 launched with an 8,192-token context length and a limited 32,768-token variant, not 128,000; the 128K window is explicitly attributed to GPT-4 Turbo, announced eight months later at DevDay in November 2023 (Source 5, openai.com), a distinction corroborated independently by Wikipedia (Source 12) and a contemporaneous dev.to timeline (Source 13). The most reliable and directly relevant sources—OpenAI's own contemporaneous announcements and technical report—unambiguously refute the claim that GPT-4 had a 128,000-token context window at its March 2023 release, making the claim false.
Reviewer 3 — The Precision Analyst
OpenAI's March 14, 2023 release material states that GPT-4 had an 8,192-token context length, with limited access to a 32,768-context variant (Sources 4 and 6); the 128K window was announced for GPT-4 Turbo in November 2023 (Source 5). The claim therefore conflates the original GPT-4 release with a later successor model and is false as worded.
Panel summary
Contemporaneous OpenAI release materials are the strongest evidence and specify an 8,192-token context window for the original GPT-4, alongside limited access to a 32,768-token variant. Independent timelines corroborate that 128K context arrived with GPT-4 Turbo in November 2023. The claim's reasoning conflates two differently timed model releases, while its central number and date are both contradicted by the primary documentation. Source quality, logic, and quantitative precision therefore all support a False verdict.