QR-02

LLM Provider & Model Comparison

Prices move weekly. This page is refreshed monthly, always confirm against the provider before you sign.
Last verified 2026-07-09

Core Principles

There is no best modelThere are only best fits for a workload, a budget, and a latency target
Route, do not upgradeA tier spread of 5x to 25x means routing beats picking a single "best" model
Cache the stable partCached input runs around 90% cheaper. Worth the effort every time
Batch anything offlineAround 50% off, where offered. Free money for non-urgent jobs
Verify before you budgetVendor pages move faster than aggregators. Trust the source, not the summary

Know The Limits read this first

This table showsPrices, context windows, and typical fit. Not a quality ranking
This table does not showBenchmarks. Leaderboards move faster than prices, and rankings depend on the eval
Never budget from this pagePrices change weekly. Confirm the official pricing page on contract day

Master Comparison Table USD per million tokens, verified 2026-07-09

ModelTierInputOutputCached InContextBest for
Anthropic (Claude)
Claude Opus 4.8Frontier5.0025.000.501MComplex reasoning, agentic coding, long-horizon autonomy
Claude Sonnet 5 *Balanced2.0010.000.201MDefault production workhorse: coding, RAG, analysis
Claude Haiku 4.5Budget1.005.000.10200KClassification, routing, extraction, moderation
Claude Fable 5Premium10.0050.001.001MWhere fewer failed loops justify the token rate
OpenAI (GPT)
GPT-5.5Frontier5.0030.000.501MHardest coding and agent tasks
GPT-5.4Balanced2.5015.000.251.05MLower-cost frontier work. Strong cost-quality balance
GPT-5.4 miniBudget0.754.500.075400KRouted production traffic below the frontier
GPT-5.4 nanoUltra0.201.250.02400KHigh-volume routing, extraction, classification
Google (Gemini)
Gemini 3.1 ProFrontier2.0012.000.201MTop-tier reasoning, multimodal, long-context
Gemini 3.5 FlashBalanced1.509.000.151MSpeed-meets-frontier with native grounding
Gemini 3 FlashBudget0.503.000.051MCost-effective general work
Gemini 2.5 Flash-LiteUltra0.100.400.011MCheapest Google tier: classification, routing at scale
DeepSeek
DeepSeek V3Ultra0.271.100.027128KLowest-cost flagship-class option for text
* Sonnet 5 introductory pricing runs to 31 Aug 2026, then reverts to $3.00 / $15.00 per million on 1 Sep 2026. Gemini Pro tiers charge roughly 2x input / 1.5x output for prompts over 200K tokens.
QR-02

LLM Provider & Model Comparison

Cost levers, model-to-job mapping, provider notes, and how to pair this with the calculator.
Last verified 2026-07-09

Cost Levers try in this order, cheapest first

LeverTypical savingHow
Trim the contextFreeDelete stale history, tighten system prompt, drop irrelevant docs. Attention dilutes anyway
Prompt cachingUp to ~90% off repeatsPut stable content (system, examples, docs) at the front, byte-identical every call
Batch processing~50% offOvernight jobs, evals, backfills. Latency for money
Tier routing5x to 25x cheaperClassify the query, route easy traffic to a budget model, keep the frontier for the hard ones
Cap output tokensVariableOutput usually costs 4x to 6x input. Bound it in the request
Move up a tierLast resortOnly after the above. A bigger model is a lever, not an answer

Which Model For Which Job

WorkloadReach for
Classification, routing, extractionUltra tier (Haiku 4.5, GPT-5.4 nano, Flash-Lite)
RAG synthesis, general chatBalanced tier (Sonnet 5, GPT-5.4, Gemini 3.5 Flash)
Agentic coding, long tasksFrontier (Opus 4.8, GPT-5.5, Gemini 3.1 Pro)
Multimodal, images, videoGemini 3.1 Pro or Opus 4.8
Long-context Q&A (500K+)Any 1M-window model; watch Gemini Pro's price step at 200K
High-volume, low latencyBudget or Ultra tier; DeepSeek V3 if privacy allows
Regulated or safety-criticalFrontier + evals + guardrails. Never the ultra tier

Provider Notes watch on the monthly refresh

Sonnet 5 price stepReverts to $3 / $15 on 1 Sep 2026. Hard-code a reminder
Claude Fable 5Access history: suspended 12 Jun 2026, restored 1 Jul 2026. Confirm before adoption
OpenAI GPT-5.6 familyLimited preview in July 2026. May go GA and displace GPT-5.5
Gemini long-contextPro tiers step up ~2x above 200K tokens
DeepSeekFigures vary widely across aggregators. Trust the official page only

Best Practices

Do
Track token spend per feature, not per model
Cache the front of your prompt aggressively
Route by query complexity, not by mood
Evaluate models on your workload, not on public benchmarks
Set hard cost caps and rate limits at deploy
Re-check pricing before every quarterly budget
Do not
Pick a single model for the whole product
Trust aggregator pricing pages. They lag
Skip caching because "the code is more complex"
Assume last month's benchmark still holds
Confuse a bigger context with better performance
Ship without a per-feature spend alert

How To Use This Page alongside the calculator and routing card

QR-09 Cost CalculatorEnter your workload, see monthly cost across every model above. Reads the same pricing file as this page
QR-15 Routing CardWhich tier for which job. Small/fast vs reasoning-heavy, when to enable extended thinking
Business Case & ROI (BCR-05, BCR-06)Both articles link back here rather than embedding prices. Never trust an inline price older than a month