Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →In a January 31, 2025 Reddit AMA, Sam Altman called DeepSeek “a very good model” and said OpenAI expected to keep producing better models—but with a smaller lead than before. He also said OpenAI was discussing a different approach to openness. Those comments signaled a tougher competitive environment, not an admission that DeepSeek had overtaken OpenAI or a promise that OpenAI would release its model weights.
What Sam Altman said about DeepSeek
Altman made the remarks during an OpenAI Reddit AMA on January 31, 2025, joined by other OpenAI leaders. His central assessment was that DeepSeek had built a strong model and that OpenAI’s advantage would be narrower than it had been in previous years.
He also said OpenAI was discussing releasing model weights and publishing more research. Altman said he personally believed the company had been “on the wrong side of history” on open source, while making clear that not everyone at OpenAI shared his view and that openness was not the company’s highest priority. This was a statement of opinion and ongoing discussion—not a release announcement, policy commitment, or timetable.
On products and pricing, Altman said OpenAI did not plan to raise the $20-per-month ChatGPT Plus price at that time and that he would like to lower it over time. He expected o3 to arrive in more than a few weeks but less than a few months, and described o3 Pro as potentially valuable to people seeking the strongest reasoning performance. These were January 2025 comments, not confirmation of current prices or product availability.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
Why DeepSeek-R1 unsettled the competitive picture
DeepSeek’s January 2025 R1 release combined a reasoning model with a technical report, model materials, distilled variants, and API pricing information. DeepSeek said some distilled models were comparable to OpenAI’s o1-mini on selected benchmarks; that was the company’s claim, not an independent finding that every variant matched o1-mini across tasks.
The disruption came from more than a single score. DeepSeek made a credible case that useful reasoning could be offered at low advertised API prices, that model weights could be deployed outside the original vendor’s hosted service, and that smaller distilled models could give developers more options. Those features challenged assumptions about the cost and access needed to compete. They do not establish that all tasks, production conditions, or total operating costs were equivalent.
“DeepSeek” also refers to distinct products and deployment choices: R1, distilled R1 variants, DeepSeek’s hosted API, and self-hosted or third-party-hosted models. Their performance, cost, terms, and operational responsibilities should not be treated as interchangeable. The R1 release materials describe the launch and DeepSeek’s own comparisons.
Did DeepSeek overtake OpenAI?
The AMA does not prove that it did. Altman’s wording suggested OpenAI still expected to produce better models, while conceding that the gap had narrowed. The AMA supplied no comprehensive benchmark comparison, independent evaluation, or audited cost comparison that would settle which company led overall.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Even a broad “which model is best?” ranking can hide the differences that matter in practice. A buyer should compare models on the actual work they need done, including:
- Reasoning, coding, and mathematical problem-solving on representative tasks
- Factuality, tool use, and structured-output reliability
- Latency, uptime, rate limits, and context-window needs
- Safety behavior and suitability for the application
- Data handling, jurisdiction, and deployment flexibility
- Total cost, including engineering and infrastructure—not just token charges
OpenAI’s o3-mini announcement emphasized reasoning and coding alongside lower cost and latency. Its system card provides evaluation details and limitations. Benchmark results can vary with prompts, settings, model versions, and test dates, so a vendor’s selected benchmark claim should not be treated as a universal ranking.
OpenAI’s immediate product response: o3-mini
OpenAI launched o3-mini on the same day as the AMA. It presented the model as a lower-cost reasoning option with adjustable reasoning effort, available through its API and ChatGPT offerings, with access and limits varying by plan. OpenAI also highlighted function calling and structured outputs for developers. At launch, o3-mini did not support vision; OpenAI directed developers needing visual reasoning to o1.
The timing made the launch an obvious part of the competitive story, but it does not show that DeepSeek caused OpenAI to create the model. The more supportable interpretation is that DeepSeek sharpened the context: cost, speed, reasoning capability, and access had become especially visible points of competition.
Recommended Free Tools
What “open source” meant in the debate
DeepSeek made weights and related materials available, offering more deployment flexibility than a proprietary model accessible only through a hosted product or API. But “open weights” and “open-source AI system” are not automatically equivalent. A system may expose weights and some code or documentation without publishing everything needed to reproduce it or independently assess it.
- Closed or proprietary model: The user accesses a hosted product or API, but cannot download the model weights.
- Open-weight model: The weights are available for download or deployment under stated terms.
- Open-source AI system: A broader, contested label that can involve code, weights, documentation, and sufficiently reproducible components.
Weights alone do not necessarily reveal training data, data-processing pipelines, complete training infrastructure, every reinforcement-learning procedure, safety filters, or all reproducibility details. Terms also matter: availability for download does not by itself establish unrestricted commercial rights.
Altman’s call for a “different open source strategy” left key questions unanswered: which model, what license, what materials, and when. It should not be read as a commitment to open-source OpenAI’s frontier models.
How the economics could affect AI buyers
DeepSeek’s importance did not depend on winning every benchmark. A buyer might accept somewhat weaker performance in exchange for lower usage costs, local deployment, customization, or less dependence on one hosted vendor. Conversely, an inexpensive model can be a poor deal if it requires more engineering, performs unreliably, or lacks suitable support and controls.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsThere are three kinds of price pressure to watch:
- Consumer subscriptions: Users may expect capable reasoning to be included in free or lower-cost plans.
- Developer APIs: Teams can compare cost per input and output token, but must also test quality, latency, limits, and integration work.
- Inference economics: Lower cost per query may encourage more usage even as revenue per query falls; whether that trade-off works depends on demand and operating costs.
Altman’s Plus comments and OpenAI’s o3-mini pricing message are historical signals from January 2025, not a current price comparison. The sources here do not establish how prices or market shares changed afterward.
Distillation claims are not proof of copying
OpenAI argued that DeepSeek may have used distillation—using answers from a stronger model to train another model. A Council on Foreign Relations discussion described the allegation and the uncertainty around its longer-term significance. The claim should remain attributed: it is not, on the evidence cited here, a settled explanation of DeepSeek’s results.
Technical validity, evidence of how a model was trained, and the legal significance of any training method are separate questions. The AMA also does not show that OpenAI copied DeepSeek. Announcing o3-mini on the same date is not evidence of shared architecture or training methods.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to compare OpenAI and DeepSeek for a real use case
For individual users
Test the tasks you actually do, then weigh answer quality against price and the features you need, such as web access, multimodal input, integrations, or higher usage limits. Avoid entering sensitive information into a service unless its data practices suit your needs. A hosted assistant may be more convenient; a model that can be deployed locally may offer more control but requires technical setup.
For developers
Run the same representative prompts through the models and record quality, latency, failure rates, token use, and the effort needed to integrate tool calling or structured outputs. Check current API terms, data retention, rate limits, regional availability, and support. Downloadable weights can enable self-hosting or third-party hosting, but shift responsibility for infrastructure, security, monitoring, updates, and scaling to your team or provider.
For businesses
Include contractual data protections, residency, auditability, security review, model-change notices, liability terms, abuse monitoring, procurement, and support in the comparison. A low token price is not the same as a low total cost of ownership, particularly when self-hosting demands GPU capacity and specialist staff.
What the AMA did—and did not—settle
Altman acknowledged that DeepSeek had narrowed the competitive gap and made OpenAI reconsider how it approached openness. He did not concede an overall technical defeat, announce a model-weight release, or establish a new pricing policy. The longer-term questions—whether OpenAI would release meaningful weights, how low-cost models would affect business economics, and how buyers would value self-hosting—were left open by the AMA.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
