Be a part of the occasion trusted by enterprise leaders for practically 20 years. VB Rework brings collectively the folks constructing actual enterprise AI technique. Study extra
Excellent news, AI builders!
OpenAI has introduced a substantial worth reduce on o3, its flagship reasoning massive language mannequin (LMM), slashing prices by a whopping 80% for each enter and output tokens.
(Recall tokens are the person numeric strings that LLMs use to characterize phrases, phrases, mathematical and coding strings, and different content material. They’re representations of the semantic constructions the mannequin has discovered by means of coaching, and in essence, are the LLMs’ native language. Most LLM suppliers provide their fashions by means of software programming interfaces or APIs that builders can construct apps atop of or plug their exterior apps into, and most LLM suppliers cost them for the privilege at a value per million tokens).
The replace positions the mannequin as a extra accessible possibility for builders looking for superior reasoning capabilities, and locations OpenAI in additional direct pricing competitors with rival fashions equivalent to Gemini 2.5 Professional from Google DeepMind, Claude Opus 4 from Anthropic, and DeepSeek’s reasoning suite.
Introduced by Altman himself on X
Sam Altman, CEO of OpenAI, confirmed the change in a put up on X highlighting that the brand new pricing is meant to encourage broader experimentation, writing: “we dropped the value of o3 by 80%!! excited to see what folks will do with it now. suppose you’ll even be pleased with o3-pro pricing for the efficiency :)”
The price of utilizing o3 is now $2 per million enter tokens and $8 per million output tokens, with an additional low cost of $0.50 per million tokens when the consumer enters data that has been “cached,” or is saved and similar to what they offered earlier than.
This marks a big discount from the earlier charges of $10 (enter) and $40 (output), as OpenAI researcher Noam Brown identified on X.
Ray Fernando, a developer and early adopter, celebrated the pricing drop in a put up writing “LFG!” quick for “let’s fucking go!”
The sentiment displays a rising enthusiasm amongst builders trying to scale their initiatives with out prohibitive mannequin entry prices.
Value comparability to different rival reasoning LLMs
The value adjustment comes at a time when AI suppliers are competing extra aggressively on each efficiency and affordability. A comparability with different main AI reasoning fashions illustrates how vital this transfer could possibly be:
- Gemini 2.5 Professional Preview, developed by Google DeepMind, expenses between $1.25 and $2.50 for enter relying on immediate dimension, and $10 to $15 for output. Whereas its integration with Google Search affords further performance, that service carries its personal value — free for the primary 1,500 requests per day, then $35 per thousand requests.
- Claude Opus 4, marketed by Anthropic as a mannequin optimized for complicated duties, is the most costly of the group, charging $15 per million enter tokens and $75 for output. Immediate caching learn and write companies come at $1.50 and $18.75 respectively, though customers can unlock a 50% low cost with batch processing.
- DeepSeek’s fashions, significantly DeepSeek-Reasoner and DeepSeek-Chat, undercut a lot of the market with aggressive low pricing. Enter tokens vary from $0.07 to $0.55 relying on caching and time of day, whereas output ranges from $1.10 to $2.19. Discounted charges throughout off-peak hours convey costs down even additional, to as little as $0.035 for cached inputs.
| Mannequin | Enter | Cached Enter | Output | Low cost Notes |
|---|---|---|---|---|
| OpenAI o3 | $2.00 (down from $10.00) | $0.50 | $8.00 (down from $40.00) | Flex Processing: $5 / $20 |
| Gemini 2.5 Professional | $1.25 – $2.50 | $0.31 – $0.625 | $10.00 – $15.00 | Greater charge applies to prompts >200k tokens |
| Claude Opus 4 | $15.00 | $1.50 (learn) / $18.75 (write) | $75.00 | 50% off with batch processing |
| DeepSeek-Chat | $0.07 (hit)$0.27 (miss) | — | $1.10 | 50% off throughout off-peak hours |
| DeepSeek-Reasoner | $0.14 (hit)$0.55 (miss) | — | $2.19 | 75% off throughout off-peak hours |
As well as, impartial third-party AI mannequin comparability and analysis group Synthetic Evaluation ran the brand new o3 by means of its suite of benchmarking exams on numerous duties, and located it value $390 to finish all of them, versus $971 for Gemini 2.5 Professional and $342 for Claude 4 Sonnet.
Narrowing the associated fee vs. intelligence hole for builders
OpenAI’s pricing transfer not solely narrows the hole with ultra-low-cost fashions like DeepSeek but in addition places downward strain on higher-priced choices like Claude Opus and Gemini Professional.
In contrast to Claude or Gemini, OpenAI’s o3 additionally now affords a flex mode for synchronous processing that expenses $5 for enter and $20 for output per million tokens, giving builders extra management over compute value and latency relying on workload kind.
o3 is at present accessible by means of the OpenAI API and Playground. Customers with balances as little as just a few {dollars} can now discover the mannequin’s full capabilities, enabling prototyping and deployment with fewer monetary obstacles.
This might significantly profit startups, analysis groups, and particular person builders who beforehand discovered higher-tier mannequin entry cost-prohibitive.
By considerably decreasing the price of its most superior reasoning mannequin, OpenAI is signaling a broader pattern within the generative AI area: premium efficiency is rapidly turning into extra inexpensive, and builders now have a rising variety of viable, economically scalable choices.
