All articles
Page 5 of 12
LLM EconomicsWhy a GPU Cloud Rate Card Is a Ceiling, Not a Price
CoreWeave's published H100 rate is $6.16 per GPU-hour, but the best committed discount it advertises lands in the same band as a no-commitment marketplace rate, which means a multi-year contract buys capacity certainty rather than savings.
LLM EconomicsH100 vs A100 Cost: Why a Faster GPU Can Still Raise Your Bill
An H100 SXM5 costs about 2.75x more per hour than an A100 80GB SXM4, so it only saves money when its throughput advantage clears that same ratio.
LLM EconomicsLatency Now Has a Price Tag: What Faster Inference Is Actually Worth
Anthropic charges exactly double for Fast Mode, which gives you a market price for latency and therefore a ceiling on what any speed optimization is worth to your business.
LLM EconomicsOpen Model Adoption vs Attention: What the Download Data Means for Your Cost Model
Hugging Face's summer 2026 data shows the models the field talks about and the models it actually runs are almost entirely different sets, and your cost model should follow the second one.
LLM EconomicsClaude Opus 5 Pricing: Why the Same $5/$25 Rate Can Still Raise Your Bill
Opus 5 kept Opus 4.8's per-token price, but thinking now runs by default, which raises the output share of a typical request and pushes your real blended rate above the sticker math.
Pricing StrategyCredit Pricing and Margin Guardrails: What Leonardo.ai's Token Model Teaches AI Founders
A credit layer works when it absorbs variable compute cost, and it only stays safe when your unlimited tiers cover models you actually own.
LLM EconomicsWhy Harness Optimization Beats Model Switching for Cutting LLM Costs
Changing the scaffolding around your model cut inference cost by an average of 40% in Writer's own research, which was often a more reliable lever than changing the model itself.
Founder GuidesHow to Cut AI Agent Costs Without Downgrading Your Product
The largest cost lever in an AI agent is no longer which model you call, it is where you draw the line between work that needs judgement and work that only needs data moved.
LLM EconomicsWhat 14X Faster Inference Does to Your AI Cost Model
Speed is now a purchasable tier rather than a property of your model, which means latency becomes a line item you choose per endpoint instead of a constraint you inherit.
LLM EconomicsAgent Memory Token Cost: Why Delivery, Not Storage, Decides Your AI Bill
Two agent-memory systems with the same lessons and the same accuracy can differ by nearly 7x in tokens per task, because the cost is set by how much context you send at inference, not by how much you keep.
LLM EconomicsDistillation Just Got 15x Cheaper. Here Is What That Does to Your AI Cost Model
A new memory-efficient training method cut peak VRAM in knowledge distillation from 85.2 GiB to 5.45 GiB at 32K context, which moves the volume at which running your own small model beats paying per token.
Pricing StrategyChatGPT Business Premium Seats: 5x the Price for 5x the Usage, and Why That Is Unusual
OpenAI priced its new Premium seat at $125 per user per month against $25 for Standard, exactly 5x for exactly 5x the usage, which is rarer in SaaS pricing than it sounds.
Pricing math, in your inbox.
One short note a week on AI pricing, token economics, and margin. No spam, unsubscribe anytime.