OpenAI is getting desperate

Summary

The video analyzes OpenAI’s decision to double usage limits for its paid plans, effective October 29, and asks whether the change makes its coding subscriptions more competitive with Anthropic’s Claude plans.

Plan changes and usage limits

• Usage limits doubled for the $100, $200, and $500 Pro tiers, as well as Business Premium.
• For the $200 tier, the increase largely restores limits that were reduced a month earlier. The $100 and $500 tiers receive a larger increase.
• The $500 plan now offers roughly 50 times the usage of the base Pro tier.
• ChatGPT and work usage now count toward the same limits as Codex. The presenter says this is unlikely to meaningfully affect most coding-heavy users, though it could matter to people who use those products extensively.
• The video does not clarify whether usage in “Dots” counts toward the same limits.

Comparing subscription value

The presenter estimates that the $200 OpenAI plan provides about $4,300 worth of API usage, while the $500 plan provides about $10,500. Claude’s $200 plan is estimated to provide about $12,000 in usage before restrictions, or about $6,000 after accounting for a 50% limit on one model. By this measure, Claude still offers more nominal value.

However, token prices alone can be misleading because models may use different numbers of tokens to complete the same task. The presenter argues that OpenAI models can be substantially cheaper per task, while Anthropic’s strongest model performs better. Subscription discounts and usage limits also narrow the apparent price gap.

The analysis accounts for frequent OpenAI usage resets—11 in the preceding 30 days. Without those resets, OpenAI’s subscription limits look less favorable. The presenter notes that reset frequency has varied significantly over time.

Blind review test

The presenter used a “council” workflow to have multiple models review the same coding changes. Across 83 comparisons:

• Opus was preferred in 66 reviews by Opus and 54 reviews by OpenAI’s model.
• Median review costs were about $1.65 for Opus and $0.36 for the OpenAI model.
• Based on the presenter’s subscription-value estimates, Claude’s $200 plan could cover roughly 205 Opus reviews, compared with about 290 reviews using the OpenAI model on ChatGPT Pro.

The OpenAI option could therefore handle more reviews, but the presenter considers Opus notably more capable.

Overall assessment

The limit increase is welcome, especially for the $100 and $500 plans, but it does not address the presenter’s main concern: OpenAI still needs a model that can compete with Anthropic’s leading Opus model.

Sponsored product feature

The sponsored segment presents WorkOS Passport as a shared sign-in and access-management layer for apps and AI agents. Agents can use connected apps through MCP servers, while organizations can apply company-wide policies governing those connections.

Reviews, guides, and tutorials

• Review and analysis: A blind comparison of model-generated code reviews, with estimates of how many reviews each subscription could cover.
• Plan comparison: An analysis of subscription limits, API-equivalent value, cost per task, and the effect of usage resets.
• Tutorials: No step-by-step tutorial is presented.

Main speakers and sources

• Theo: Video presenter, discussing OpenAI and Anthropic plan details and his own usage comparisons.
• Tibo: Cited for the OpenAI usage-limit announcement.
• Adam Holter: Credited for subscription-value and usage-comparison charts.
• WorkOS Passport: Featured in the sponsored segment.