AI ToolsOpenAI has unveiled GPT-5 — its most powerful language model yet. We break down the key improvements and differences from GPT-4.
OpenAI shipped GPT-5, and this isn't the usual incremental bump with a "now even better" press release. The model is live for all ChatGPT Plus subscribers, and after a week of testing — the jump from GPT-4 is real and obvious.
A year and a half. That's how long the AI world waited after GPT-4 shook up the industry. In the meantime, Anthropic released Claude 4, Google refined Gemini 2.0, and Chinese labs started closing the gap fast. OpenAI needed a big swing. Did they connect? Let's look at the specifics.
The context window ballooned to 256K tokens — roughly 500 pages of text. Load an entire book, a full codebase, or months of email threads into a single conversation. For anyone who analyzes documents or writes code for a living, this is transformative.
Speed doubled compared to GPT-4. The improvement is most noticeable on long outputs — the model no longer sits there "thinking" for 30 seconds before responding.
Reasoning took a serious step up. Multi-step logic problems, math, formal proofs — GPT-5 handles them with a confidence that GPT-4 never had. Sam Altman claims internal benchmarks show "a leap comparable to the GPT-3 to GPT-4 transition." Bold claim, but our testing backs it up on coding and analysis tasks.
Native multimodality means text, images, audio, and video all flow through one conversation. And hallucinations dropped noticeably. The model now says "I'm not sure" instead of inventing facts with full confidence. Honestly? That's the improvement we're most excited about.
Independent benchmarks put GPT-5 ahead in reasoning and code generation. Claude 4 still edges it out on long-context work — which is ironic, since GPT-5 technically has a bigger window. Gemini 2.0 wins on multimodal tasks involving video, which makes sense given Google's YouTube infrastructure advantage.
For users weighing access options: ChatGPT still requires workarounds in some regions. Claude has similar restrictions. GigaChat is the easiest to access in Russia but can't match the raw capability yet. Pick based on what you actually need, not benchmark scores.
ChatGPT Plus stays at $20/month but comes with message caps. Want unlimited? ChatGPT Pro at $200/month — a steep ask for individuals. API pricing lands at $15 per million input tokens and $60 per million output tokens. For businesses, that's cheaper than a junior hire. For hobbyists comparing against Claude's API, it's pricier.
GPT-5 isn't an incremental upgrade. It's a genuine leap. If you use ChatGPT daily — writing code, drafting copy, analyzing data — the upgrade justifies itself within the first few days. Text quality is sharper, code output is more reliable, and the expanded context unlocks workflows that simply weren't possible before.
The flip side: the gap between free and paid has never been wider. OpenAI is making it very clear they want your credit card. And at $200/month for Pro, they want it badly.
Midjourney has launched a web interface with an image editor, accessible without Discord.
ReadSuno has released version 4 — music generation has reached studio quality.
ReadAI code editor Cursor has surpassed GitHub Copilot in active users.
Read