Every model you trusted last quarter just became the second-best option available.

The model-switching tax is real and expensive

Professionals have been routing tasks across three or four different AI tools because no single model handled code, math, vision, and writing at the same level. That fragmented workflow costs time and introduces errors every time context has to be rebuilt from scratch.

One prompt box now covers what four tools used to split

Introducing GPT accepts text, images, and complex reasoning tasks in a single interface and returns outputs that previously required model-switching mid-project. You write or paste your task, attach what you need, and the model handles coding, mathematical reasoning, health-related analysis, and visual interpretation inside one session. The output is ready to copy into your actual deliverable.

The people who feel this fastest work in high-stakes outputs

• Software engineers debugging and writing production code who need a model that stops hallucinating function signatures
• Financial analysts running quantitative work who previously hit the ceiling on math accuracy with GPT-4o
• Medical and health researchers who need visual perception and document comprehension treated with equal weight

The common thread is professionals whose work gets audited, meaning errors have real consequences.

The intelligence gap between last year and today just became impossible to ignore

OpenAI’s internal benchmarks show GPT-5 outperforming every prior OpenAI model across every tested domain, and Anthropic’s Claude 3.7 Sonnet held the reasoning crown for much of early 2025. If that competitive pressure forced OpenAI to ship a genuinely cross-domain model rather than a narrow improvement, the pace of capability releases this year is about to accelerate across every major lab.

What you can actually do with it today

• Write, debug, and explain production code without switching to a specialized model
• Solve multi-step math and quantitative problems in a single session
• Analyze charts, screenshots, or medical images alongside written context
• Draft long-form content that holds argument structure across thousands of words

GPT-5 is available now through ChatGPT and the OpenAI API.

Pricing: Pricing tiers have not been fully published for all access levels — check the OpenAI site directly for current API and subscription costs.

Honest tradeoff: A model this capable raises the cost of a wrong answer — users who over-trust outputs in high-stakes domains without verification are taking on more risk, not less.

Alternatives: Anthropic’s Claude 3.7 Sonnet remains the strongest direct competitor for reasoning-heavy tasks. Google’s Gemini 1.5 Pro covers multimodal inputs and long context if document length is your primary constraint.

The era of routing tasks across multiple AI models is ending

The shift toward single-model workflows is happening faster than most teams have planned for, and the Friday newsletter is where we track exactly which tools are making that transition real. We cover tools like this every Friday — subscribe here and we’ll send the best ones straight to you.