"$200 and Still Discriminated": Claude's Never-Ending Pricing Controversy

On February 5, Anthropic unveiled its most powerful model, Claude Opus 4.6. With top scores on Terminal-Bench 2.0, a 1M token context window (beta), and Adaptive Thinking, the technical achievement is undeniable. But behind the impressive launch lies an uncomfortable truth: $200 subscribers still hit usage limits, the 1M context is API-only, and third-party tools are blocked.

1. Fast Mode: 6x Price After Promo Ends

Opus 4.6 introduced a 'Fast Mode' for Claude Code—same model, faster output. The catch is pricing: '$15/$75 per Mtok (50% off through Feb 16).' After the promo ends, the full price hits $30/$150—6x the standard mode ($5/$25) and 2x the already expensive Opus 4.1 era pricing.

Model/Mode	Input (Mtok)	Output (Mtok)	Note
Opus 4/4.1	$15	$75	Previous gen (highest)
Opus 4.5	$5	$25	67% reduction
Opus 4.6 Standard	$5	$25	Price maintained
Opus 4.6 Fast (current)	$15	$75	50% promo (~02.16)
Opus 4.6 Fast (full)	$30	$150	After promo ends

The pricing goodwill built by Opus 4.5's 67% reduction has vanished behind a 'speed' premium. Critics compare it to mobile game monetization—free taste, paid meal.

2. Context Window: $200 Subscribers Are Second-Class

The 1M token context window is Opus 4.6's most impressive spec, also supported by Sonnet 4.5 and Sonnet 4 in beta. But it's restricted to API Tier 4+ organizations, while $200 Max subscribers remain locked at 200K tokens.

Access Method	Context Window	Cost
API (Tier 4+ org)	1M tokens (beta)	Pay-per-use
Claude Code pay-per-use	1M tokens (beta)	Pay-per-use
Max 20x subscription	200K tokens	$200/mo
Max 5x subscription	200K tokens	$100/mo
Pro subscription	200K tokens	$20/mo ($17 annual)

While $200/month subscribers are capped at 200K tokens, API developers freely use 5x that. It's a paradox: those who pay the most get the least features.

3. Usage Limits: A Black Box

Anthropic doesn't disclose exact message limits per plan. The pricing page only says 'Usage limits apply'—no specifics on how many requests per 5-hour window.

GitHub Issue #16157 shows the scale of frustration. Titled 'Instantly hitting usage limits with Max subscription,' it has 1,175+ comments and 510+ reactions, making it the most-reacted issue in the repository's history.

In July 2025, weekly limits were introduced without notice. In late December, a 'holiday bonus' doubled limits, but when they reverted in January, users suspected limits were reduced below the original level. The Register reported an estimated 60% token limit reduction, while Anthropic simply called it a 'resumption of normal limits.'

"33% of my Max 5-hour usage was consumed in 27 minutes. I was running a single session. This is insane." — GitHub Issue #22435 (with mitmproxy data)

4. Third-Party Blocking and Developer Exodus

That's not all. On January 9, 2026, Anthropic blocked all third-party coding tools from Claude Max API access. OpenCode (56k+ stars), Cline, and Roo Code went dark simultaneously. The trigger was the 'Ralph Wiggum' technique—a 5-line bash script that kept Claude Code running autonomously overnight.

"Seems very customer hostile to block third-party agents from using a paid subscription." — DHH (Ruby on Rails creator)

The backlash was fierce. OpenCode added ChatGPT support within hours, and OpenAI extended support to OpenHands, RooCode, and other open-source tools. An even more painful comparison emerged—GitHub Copilot offers Claude Sonnet 4.5 'unlimited' for just $10/month.

5. Same $200, Different Worlds: OpenAI Comparison

The starkest contrast is the separated limits policy. In ChatGPT, general conversations and Codex coding have independent usage limits. You can use Codex all day without affecting chat limits. Anthropic unifies claude.ai and Claude Code into a single limit—coding eats into your conversation quota. Hitting 'limit reached' on a simple question after coding is especially frustrating for $200 subscribers.

Comparison	Claude Max 20x ($200)	ChatGPT Pro ($200)
Chat usage	Unified with Claude Code	Unlimited* (separate from coding)
Coding tool usage	Unified with chat	Codex separate limit (6x + priority)
Top model	Opus 4.6	GPT-5.2 Pro (exclusive)
Context window	200K tokens (1M is API Tier 4 only)	400K tokens
Third-party tools	Claude Code only	GitHub/Slack/Linear integration
Over limit	Slowdown or buy credits	Buy additional credits
Mid-tier option	Max 5x ($100)	None (Plus $20 → Pro $200)

Transparency differs too. OpenAI publishes Codex usage limits by plan.

Codex Usage (per 5hr)	Plus ($20/mo)	Pro ($200/mo)	Ratio
Local tasks	45-225	300-1,500	~6x
Cloud tasks	10-60	50-400	~6x
Code reviews (weekly)	10-25	100-250	~10x

Not exact numbers, but enough to set expectations. Pro gets priority processing, credit top-ups when over limit, and a lighter model option (4x more limit). Anthropic offers undisclosed limits, third-party blocks, and unified quotas—the exact opposite.

Final Thoughts: Is the #1 Crown Enough?

Opus 4.6's technical achievement is undeniable—Terminal-Bench 2.0 record, 144 Elo points ahead of GPT-5.2 on GDPval-AA. API pricing at $5/$25 is moving in the right direction.

But opaque limits, subscriber-API feature gaps, third-party blocking, and 6x post-promo pricing are building a narrative: 'great technology, but no respect for customers.'

Anthropic mocked OpenAI's ad commercialization as 'Betrayal' in its Super Bowl ad. But for $200 subscribers struggling with limits, charging premium prices without ads while restricting usage isn't clearly the better deal. Sam Altman's retort—'a company that serves an expensive product to rich people'—isn't entirely wrong.

OpenAI is expanding with generous limits and ecosystem integration, while Google DeepMind pushes value with 2M token context. As both giants rapidly close the technology gap, Anthropic stands alone in maintaining a 'pay more, get less' user experience. The era where being #1 in technology shields you from policy criticism may already be over.