Changelog

What changed,
and what it cost.

Every product update and price movement — explained in words, not API docs. Because a 400% price increase means more when you see it as a novel going from $0.35 to $1.33.

May 2026 Price Change

Gemini Flash 3.5 quietly got 5× more expensive

Google updated pricing on their mid-tier Gemini model. Silver (Flash-Lite) stayed the same. Gold (Flash 3.5) did not.

+400%Flash 3.5 input
+260%Flash 3.5 output
—Flash-Lite (unchanged)
Rate changes per million tokens
Model Input before Input after Change
Flash-Lite 2.5 · Silver $0.10 $0.10 —
Flash 3.5 · Gold $0.30 $1.50 +400% ↑
🧙 The Hobbit — 95,356 words — what it now costs to process
Model Before After Difference
Flash-Lite 2.5 · Silver $0.06 $0.06 —
Flash 3.5 · Gold $0.35 $1.33 +$0.98 · +380% ↑
Gap between tiers $0.29 $1.27 4.4× wider
The irony
TokenScale launched saying "The Hobbit = $0.06 on Gemini Flash-Lite." That's still true. But the next model up — Flash 3.5 — had its Hobbit cost jump from $0.35 to $1.33 almost immediately after launch. The gap between Silver and Gold went from $0.29 to $1.27 overnight. If you were building on Flash 3.5, you just got a 380% bill increase with no warning. This is exactly what TokenScale exists to catch.
What this means practically
If you're sending novel-length context (100K tokens) to Gemini, Flash-Lite is still the smart choice — unchanged, and still the cheapest full-context model in the comparison. Flash 3.5 now costs 22× more for the same input. The tier gap matters more than ever.
20 May 2026 v2.2 · Launch Day

We caught a pricing error on our own launch morning

Hours before posting to Hacker News, we found that our hero number was quoting input cost only. Here's what we fixed — and why it made for a better story.

The correction
What we said What it actually was Corrected to
The Hobbit on Gemini Flash $0.04 (input only) $0.06 (total)
Input cost $0.01 $0.01
Output cost missing $0.05
Correct total $0.04 ✗ $0.06 ✓
The lesson
The very thing TokenScale is built to prevent — quoting only input cost and missing output — was in our own marketing copy. Caught and corrected before the HN post went live. The distinction between input and output pricing is more important than it looks. Output tokens are usually 3–5× more expensive per token, and most real conversations generate far more output than people expect.
19 May 2026 v2.1 · Pre-Launch Fix

Quiz button crash — caught one day before launch

The "Which model should I use?" quiz was silently failing on first run. A missing DOM element meant the result screen crashed before anyone could see a recommendation.

Why this matters
The quiz is the main personalisation hook — it routes new visitors to the right provider before they see pricing. A silent crash on the most common recommendation (Anthropic) would have cost us conversions on HN launch day without us ever knowing.
18 May 2026 v2.0 · Launch

TokenScale ships — 16 providers, one page

The first public version. A single HTML file, no backend, no sign-up. Pricing for 16 AI providers expressed in content you recognise.

Pricing spread at launch — mid-tier input cost per million tokens
Provider Model $/M input Hobbit cost
DeepSeekV4 Flash$0.14$0.02
GroqLlama 3.1 8B$0.05$0.01
GeminiFlash-Lite 2.5$0.10$0.06
MistralSmall$0.10$0.01
AnthropicClaude Haiku 4.5$1.00$0.13
OpenAIGPT-5.4$2.50$0.32
OpenAIGPT-5.5$5.00$0.63
Spread cheapest → most expensive35× apart
The insight that launched this
"$5 per million tokens" tells you nothing. "The Hobbit costs $0.06 on Gemini Flash-Lite, and $0.63 on GPT-5.5" tells you everything. TokenScale was built to make that translation automatic — for any content size, across all 16 providers, updated nightly.