Pricing changes, tokenizer quirks, and the optimizations we keep finding in production LLM deployments.
July 18, 2026 — guide.
Every major model's input and output token pricing in one normalized table, cheapest to most expensive — from the $0.10 budget floor to $50 flagship output. Plus how to read AI token pricing and why the cheapest model changes with your input/output mix. See the table →
July 18, 2026 — guide.
What a single token actually costs — from $0.0000001 on budget input to $0.000025 on flagship output — why input and output differ, and a worked example turning a per-token rate into a real monthly bill. Read the guide →
July 18, 2026 — guide.
What a token cost calculator does, how to use one in four steps, and how to read the result so you know an LLM's price before you ship — not after the invoice. Read the guide →
July 18, 2026 — guide.
Every LLM API charges per token — not per word, not per request. This guide explains what a token is, how prices are quoted per million tokens, why output costs about 5× input, and where the market sits in 2026 from the $0.10 budget floor to $25 flagship output. With a worked example so you can estimate your own bill. Read the full guide →
July 18, 2026 — guide.
Every Claude model's per-million rate in one place — Opus 4.8 ($5/$25), Opus 4.8 Fast and Fable 5 ($10/$50), Haiku 4.5 ($1/$5) — plus the two discounts that actually move the bill: prompt caching (~10% on cached input) and the Batch API (flat 50% off). With a worked example that cuts a $1,250 bill toward $700. Read the full breakdown →
July 18, 2026 — guide.
The exact formula for a Claude API cost estimate, step by step: count your input and output tokens, apply the per-million rate, scale to a monthly total. Includes token rules of thumb, an Opus-vs-Haiku comparison, and a free calculator that does all three steps for any Claude model. Read the guide →
June 19, 2026 — guide.
"Which LLM API is cheapest?" is the wrong question — the right one is cheapest for what you're actually doing. Output costs roughly 5× input on every frontier model, so the winner flips between chat, RAG, content, batch and agents. This guide ranks the cheapest option for each workload using live per-token pricing, then lists the five rules that beat any model choice. Read the full guide →
June 19, 2026 — guide.
We ran four real production applications side-by-side on GPT-5.5 and Claude Opus 4.8 for a full billing cycle, with identical prompts and matched user traffic. Total spend over 30 days: $4,200 on GPT-5.5, $3,650 on Claude Opus 4.8 — about 13% less on Claude, driven by its lower output rate. Per-call quality was statistically tied on three of four apps. Read the full breakdown →
June 19, 2026 — guide.
If you run an LLM application that serves Russian-speaking users, you are probably paying more than you should — and it's the tokenizer, not the price card. Across 200 prompts, Russian text cost 2.1×–2.6× more tokens than English on GPT-5.5, 1.9×–2.4× on Claude Opus 4.8, and 1.4×–1.7× on Gemini 3.5 Flash. Here's why, the exact multipliers, and four ways to cut it. Read the full guide →
June 19, 2026 — guide.
Prompt caching is the single biggest cost lever in LLM applications, and almost every team we audit is leaving 30–60% of the savings on the table. Anthropic uses explicit cache_control breakpoints, OpenAI auto-caches prefixes over ~1,024 tokens, Google requires an explicit context-caching API. This cookbook gives the working syntax for each, the minimums that actually save money, and the gotchas that quietly waste the discount. Read the cookbook →
June 10, 2026 — changelog.
Anthropic released Claude Opus 4.8 on May 28, 2026, and the headline is that the price card did not move: still $5 per million input tokens, $25 per million output tokens, same as Opus 4.7. The real news is buried in the supporting tiers. Opus 4.8 Fast — the latency-optimized variant that streams 2.5× faster than standard — now lists at $10/$50 per million tokens, which is roughly 3× cheaper than the equivalent Fast Mode on Opus 4.7 ($30/$150). For applications where a user is waiting on the response (real-time chat, voice agents, IDE completions), that delta changes the equation significantly. Anthropic also reports that Opus 4.8 is "around four times less likely than its predecessor" to produce certain failure modes — we are running our own production-traffic comparison and will publish numbers next week.
June 10, 2026 — changelog.
Anthropic followed Opus 4.8 with Claude Fable 5 on June 9, 2026. It is the first generally available Mythos-class model — Anthropic's new tier sitting above the Opus line for the hardest long-horizon work. Pricing matches Opus 4.8 Fast at $10/$50 per million tokens. The benchmark gap that justifies the upgrade is real on coding tasks: Anthropic reports 80.3% on SWE-Bench Pro versus 69.2% for Opus 4.8, with the difference concentrated on multi-file refactors and multi-repo migrations. For most application work, Opus 4.8 remains the better economic choice; Fable 5 is targeted at autonomous coding agents that run for hours without supervision and at multi-repo refactor jobs where the per-run cost is small compared to the engineer time saved.
Draft — coming soon.
RAG workloads have a specific cost profile: long inputs (retrieved chunks + question), short outputs (answer). The cost-per-call equation strongly favors models with low input rates. Currently the three contenders are Gemini 2.5 Flash-Lite ($0.10/$0.40), GPT-4.1 Nano ($0.10/$0.40), and DeepSeek V3.2 ($0.14/$0.28). We tested all three on 1,000 real customer support queries. Spoiler: the cheapest option lost on quality enough that the slightly more expensive one was the better economic choice. Full results next month.