/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

OpenAI's o3-mini costs $1.10 per 1M input tokens and $4.40 per 1M output tokens, cheaper than GPT-4o, which costs $2.50 and $10, and o1, which costs $15 and $60

Simon Willison's Weblog : X: @simonw and @daniel_mac8 . Forums: Hacker News X: Simon Willison / @simonw : Has anyone seen anything interesting done with that increased output limit yet? o1 has 100,000 as well, o1-mini is 65,536 and o1-preview was only 32,768 (Reasoning tokens come out of that same budget, but I imagine there are prompts that can reason briefly and then output long) Dan Mac / @daniel_mac8 : the fact that o3-mini has a 100,000 token output length, and that output length is so much larger in comparison to previous models lends a lot of credence to Ben's idea that the ‘o’ series reasoning models are best seen as ‘brief’ generators rather than chatbots [image] Forums: Hacker News : Notes on OpenAI o3-mini

Simon Willison's Weblog

Context & Ripple Effects

OpenAI’s reasoning line had previously been framed as a cost-and-performance trade-off relative to GPT-4o, rather than a simple upgrade path, in coverage of o1’s reasoning trade-offs.

The listed o3-mini rates sharpen that positioning: the newer reasoning model is priced below both GPT-4o and o1 on input and output tokens. Related launch coverage also characterized o3-mini as a faster, lower-cost model near o1-class capability at launch.

First-order effects

  • For API buyers, o3-mini’s stated $1.10 input and $4.40 output rates lower the direct token bill versus the listed GPT-4o and o1 rates.
  • OpenAI now has a lower-priced reasoning option, giving customers a clearer cost-based choice between o3-mini, GPT-4o, and o1.

Second-order effects

  • Teams that had treated reasoning as too costly for token-heavy workflows can reassess model routing, especially where output-token charges materially affect total spend.
  • The pricing gap increases pressure on competing model offerings to justify their own price-performance position; earlier low-cost models had already pushed token pricing downward in the GPT-4o mini comparison.

Third-order effects

  • If lower-priced reasoning models continue to approach higher-tier capability, model selection will shift further from a single flagship choice toward workload-level routing based on effective cost per useful result.
  • The market may increasingly segment between economical reasoning tiers and premium, compute-intensive tiers, with published token rates serving as a more visible product differentiator.

The trend: This is one data point in the move from premium reasoning models toward tiered inference pricing that makes advanced reasoning viable for more production workloads.

Discussion

  • @simonw Simon Willison on x
    Has anyone seen anything interesting done with that increased output limit yet? o1 has 100,000 as well, o1-mini is 65,536 and o1-preview was only 32,768 (Reasoning tokens come out of that same budget, but I imagine there are prompts that can reason briefly and then output long)
  • @daniel_mac8 Dan Mac on x
    the fact that o3-mini has a 100,000 token output length, and that output length is so much larger in comparison to previous models lends a lot of credence to Ben's idea that the ‘o’ series reasoning models are best seen as ‘brief’ generators rather than chatbots [image]