# llmhosting.ai

> Live comparison of LLM API pricing and GPU rental rates, refreshed daily from provider
> APIs (LiteLLM, OpenRouter, Helicone, Vast.ai, RunPod). LLM prices 2026-10-14,
> GPU quotes 2026-10-14. Free JSON APIs below. Attribution link appreciated.

## Data APIs (free, CORS-enabled)

- [LLM API prices](https://llmhosting.ai/api/prices.json): cheapest per-provider input/output price per 1M tokens for 300 models
- [GPU rental prices](https://llmhosting.ai/api/gpu-prices.json): live $/hr quotes per GPU per provider and tier
- [GPU price history](https://llmhosting.ai/api/gpu-history.json): daily $/hr snapshots since 2026-07-06, append-only
- [API docs](https://llmhosting.ai/api-access): terms and field reference

## Compare

- [Model API prices](https://llmhosting.ai/models): per-model pages with every provider's price
- [GPU rental](https://llmhosting.ai/gpus): per-GPU pages with live provider quotes and history
- [Providers](https://llmhosting.ai/providers): provider profiles
- [Breakeven calculator](https://llmhosting.ai/calculator): API vs self-hosted GPU math
- [Price trends](https://llmhosting.ai/trends): GPU price history charts

## Popular model pages

- [GPT-OSS 120B](https://llmhosting.ai/models/gpt-oss-120b)
- [DeepSeek V4 Flash](https://llmhosting.ai/models/deepseek-v4-flash)
- [GLM-5.2](https://llmhosting.ai/models/glm-5-2)
- [Llama 3.3 70B](https://llmhosting.ai/models/llama-3-3-70b)
- [DeepSeek V4 Pro](https://llmhosting.ai/models/deepseek-v4-pro)
- [DeepSeek R1](https://llmhosting.ai/models/deepseek-r1)
- [DeepSeek V3](https://llmhosting.ai/models/deepseek-v3)
- [Kimi K2.7 Code](https://llmhosting.ai/models/kimi-k2-7-code)
- [GPT-OSS 20B](https://llmhosting.ai/models/gpt-oss-20b)
- [GLM 5.3](https://llmhosting.ai/models/glm-5-3)
- [Kimi K2.6](https://llmhosting.ai/models/kimi-k2-6)
- [Llama 3.1 8B](https://llmhosting.ai/models/llama-3-1-8b)

## Popular GPU pages

- [A6000](https://llmhosting.ai/gpus/a6000)
- [L40S](https://llmhosting.ai/gpus/l40s)
- [6000 Ada](https://llmhosting.ai/gpus/6000-ada)
- [RTX PRO 6000](https://llmhosting.ai/gpus/rtx-pro-6000)
- [A100 SXM 80GB](https://llmhosting.ai/gpus/a100-sxm-80gb)
- [H100 SXM](https://llmhosting.ai/gpus/h100-sxm)
- [H200](https://llmhosting.ai/gpus/h200)
- [B200](https://llmhosting.ai/gpus/b200)

## Guides

- [IBM Granite 4.2 8B: the token volume where self-hosting beats the API](https://llmhosting.ai/guides/granite-4-2-8b-selfhost-breakeven): How to calculate the monthly token volume where renting a GPU for Granite 4.2 8B beats its API rate, and why the obvious GPU choice is usually the wrong one.
- [DeepSeek V4.1 Flash vs V4 Flash: is the upgrade worth the price delta?](https://llmhosting.ai/guides/deepseek-v4-1-flash-upgrade-cost): DeepSeek V4.1 Flash has early provider pricing, but the live data on V4 Flash shows why switching today is premature.
- [Ling-3.0 Flash VL: what vision input actually adds to your token bill](https://llmhosting.ai/guides/ling-3-flash-vl-vision-cost): Vision-language token overhead determines whether a VL model beats a text-only OCR pipeline on document-heavy jobs, far more than the posted per-token rate.
- [GPT-Astra, Sol, Terra, and Luna: four new tiers, priced side by side](https://llmhosting.ai/guides/gpt-astra-sol-terra-luna-pricing): A method for ranking four same-week model launches by real per-token rates instead of their names, applied to a coding-agent workload at volume.
- [DeepSeek Pro vs DeepSeek Flash: pricing the newest two-tier split](https://llmhosting.ai/guides/deepseek-pro-vs-deepseek-flash): Compares DeepSeek V4 Pro and V4 Flash on params, provider spread, and the token volume where the Pro premium stops paying for itself.
- [Pareto just got priced: where it lands on the cost/performance frontier](https://llmhosting.ai/guides/pareto-model-pricing-debut): A worked token-mix example shows the exact output-token share where Pareto's input discount against GLM-4.5 Air disappears.
- [GLM, GLM Flash, or GLM-5.3 Flash: pricing Zhipu's three live tiers](https://llmhosting.ai/guides/glm-lineup-pricing-comparison): Zhipu's GLM-5.2 and GLM-4.7 Flash both carry live per-token rates today; GLM-5.3 Flash carries none, and that gap should shape your routing call.
- [Hunyuan MT2's three sizes: a VRAM ladder from 1.8B to 30B-A3B](https://llmhosting.ai/guides/hy-mt2-family-vram-ladder): Hunyuan MT2 ships as 1.8B, 7B, and 30B-A3B, and the MoE tier needs a bigger card than its active-parameter count implies.
- [DeepSeek V4 Flash Vision: what multimodal input actually adds to your bill](https://llmhosting.ai/guides/deepseek-v4-flash-vision-cost): DeepSeek V4 Flash Vision has no live pricing row yet, so this piece prices the real text-only model and the VRAM math for adding a vision head.
- [Tencent's Hy4 just got priced: where it lands against the Hunyuan lineup](https://llmhosting.ai/guides/hy4-preview-pricing-debut): Hy4 has no live pricing row yet, so this guide shows how to bracket a new flagship MoE model's cost using param math and tracked analogs before a real quote exists.
- [GLM-5.3 Flash Batch vs standard: is Zhipu's discount as good as Gemini's?](https://llmhosting.ai/guides/glm-5-3-flash-batch-savings): Compares Zhipu's GLM Flash pricing tier against Gemini's published batch discount using live catalog data and VRAM math.
- [Claude Fable 5.1 Batch: how much of the discount survives real workloads](https://llmhosting.ai/guides/claude-fable-5-1-batch-economics): How the batch discount on Claude Fable 5.1 shrinks once queue delay, retries, and cache-hit timing enter the actual bill.
- [Qwen3.8 Max vs Qwen3.8 Flash: which tier actually fits your workload's budget](https://llmhosting.ai/guides/qwen3-8-max-vs-qwen3-8-flash): Active-parameter math and provider spread data show when Qwen3.8 Max's reasoning depth is worth its multiplier and when Flash's discount gets eaten by retries.
- [Gemini 3.7 Flash just launched: is upgrading from 3.6 worth the price delta?](https://llmhosting.ai/guides/gemini-3-7-flash-generation-upgrade-cost): A framework for deciding whether to migrate to a new Flash-tier release, using live per-token rate patterns instead of vendor release notes.
- [Nemotron 3.5 Lightning: can you actually run it cheaper than the API?](https://llmhosting.ai/guides/nemotron-3-5-lightning-local-cost): Nemotron 3.5 Lightning has no live pricing row yet, so the real cost comparison runs through its closest tracked analog and the VRAM math behind it.
- [Seed 2.0 Code vs KAT-Coder: pricing two purpose-built coding models](https://llmhosting.ai/guides/seed-2-0-code-vs-kat-coder-pricing): Compares Seed 2.0 Code and KAT-Coder against priced open-weight coding models to estimate real cost when no metered API rate is published.
- [Grok 4.6 in agent loops: how its token bill compares to Claude Opus 5](https://llmhosting.ai/guides/grok-4-6-agent-cost-vs-opus5): A worked agent-loop example shows that comparing Grok 4.6 and Claude Opus 5 on listed per-token rates misses the real cost driver: turns to completion.
- [Qwen3.8-2.4T-A95B: the VRAM math behind a trillion-plus-parameter MoE](https://llmhosting.ai/guides/qwen3-8-2-4t-a95b-vram-breakeven): The VRAM and cluster math for self-hosting Qwen3.8-2.4T-A95B, and the token volume where a hosted API price would need to land to compete.
- [Poolside's Laguna S-2.1 just got priced: how it stacks up against the incumbents](https://llmhosting.ai/guides/laguna-s-2-1-price-debut): Laguna S-2.1's listed API rates against Claude Opus 5 and GPT-5.6 Luna Pro show whether the new entrant undercuts incumbents or just adds another tab to compare.
- [The cheapest capable small model right now: Qwen3 Flash vs Gemini 3.6 Flash vs Flash-Lite](https://llmhosting.ai/guides/cheapest-small-model-july-2026): Ranking Qwen3 Flash, Gemini 3.6 Flash, and Gemini 3.5 Flash-Lite by real per-request cost using live provider data.
- [KAT-Coder Air vs Pro v2.5: pricing the two tiers against real agent workloads](https://llmhosting.ai/guides/kat-coder-v2-5-air-vs-pro): Shows when KAT-Coder Air's discount over Pro v2.5 survives an input-heavy coding agent workload and when it collapses to almost nothing.
- [Kimi K2.7 Is Trending: The Token Volume Where Self-Hosting Beats the API](https://llmhosting.ai/guides/kimi-k3-selfhost-breakeven): A memory and compute breakdown of Kimi K2.7 Code's footprint pinpoints the monthly token volume where renting B200 or H200 capacity undercuts the hosted API.
- [Gemini 3.6 Flash Batch vs standard: how much latency you trade for the discount](https://llmhosting.ai/guides/gemini-batch-pricing-payoff): Shows when Gemini 3.6 Flash and 3.5 Flash-Lite's batch discount pays for the multi-hour turnaround delay and when it doesn't.
- [Claude Opus 5 vs Opus 5 Fast: what the 'fast' tier actually costs you](https://llmhosting.ai/guides/claude-opus-5-fast-vs-standard): How to check whether Claude Opus 5 Fast's lower per-token rate actually lowers total agent cost once retry rates are counted, using live pricing and open-model analogs.
- [Multi-GPU inference: what tensor parallelism really costs](https://llmhosting.ai/guides/multi-gpu-tensor-parallel-costs): Explains how interconnect tier determines whether tensor parallelism scales linearly or falls off a cliff.
- [Renting from GPU marketplaces without getting burned](https://llmhosting.ai/guides/gpu-marketplace-safety-checklist): How to vet marketplace GPU listings for bandwidth, isolation, and hidden costs before you commit rental hours to one.
- [vLLM vs Ollama: prototyping is not production](https://llmhosting.ai/guides/vllm-vs-ollama-production): Ollama and vLLM handle concurrent requests so differently that the choice determines whether your GPU bill triples once real traffic arrives.
- [Spot and interruptible GPU rentals: when the discount is worth it](https://llmhosting.ai/guides/spot-vs-on-demand-gpu-rental): Community and spot GPU tiers run 70%+ cheaper than secure rentals on identical silicon, but the payoff depends on whether your workload survives interruption.
- [GLM-5.2, Kimi K2.7, MiniMax M3: the 'want to self-host but probably shouldn't' tier](https://llmhosting.ai/guides/big-moe-selfhost-tier): VRAM math for three frontier open-weight models shows they require multi-card GPU clusters to serve.
- [The local LLM hardware ladder, mid-2026](https://llmhosting.ai/guides/local-llm-hardware-ladder-2026): Used 24GB cards, 128GB unified memory boxes, Macs, and the RTX PRO 6000 96GB each cap out differently, and renting the same silicon is the rung most buyers skip.
- [The RAM price shock broke 2025's self-hosting math](https://llmhosting.ai/guides/ram-price-shock-buy-vs-rent): DDR5 price inflation raised the cost basis for CPU-offload MoE rigs, moving the buy-versus-rent breakeven toward renting by a wide margin.
- [NVFP4 on Blackwell: when the fancy quant actually pays](https://llmhosting.ai/guides/nvfp4-blackwell-quant-economics): NVFP4 only speeds up inference on Blackwell tensor cores; on Hopper and Ada silicon it shrinks VRAM but leaves throughput untouched.
- [DSpark made self-hosting DeepSeek V4 Flash faster, and the API price didn't move](https://llmhosting.ai/guides/dspark-selfhost-breakeven): DeepSeek's DSpark speculative decoding lowers self-hosted V4 Flash serving cost while API pricing holds, moving the rent-vs-API breakeven point.
- [What a 1M-token context actually costs (spoiler: a huge spread)](https://llmhosting.ai/guides/million-token-context-economics): How to price filling a 1M-token context window across providers and self-hosted GPUs, and when compacting context beats paying for either.
- [Total params pay the memory bill, active params pay the compute bill](https://llmhosting.ai/guides/moe-active-param-economics): Explains why sparse MoE models split into two separate cost drivers, and why that split favors A3B-class models for self-hosting over dense 70B.
- [The flat-rate agent era is over: subscription vs API vs open models, priced honestly](https://llmhosting.ai/guides/subscription-vs-api-vs-open-for-agents): How to size agent-token volume against subscription credits, raw API, open-weight models, and self-hosted GPUs now that flat-rate plans stop covering programmatic use.
- [Sticker price lies: cache-hit rates decide what you actually pay](https://llmhosting.ai/guides/cache-effective-pricing): Explains how to calculate the effective per-token price of an agent workload once cache-hit rate, write premiums, and TTLs are factored in.
- [Why your coding agent costs 10-100x what chat does](https://llmhosting.ai/guides/agent-token-economics): Agent loops rebill the growing transcript on every tool call, so cost per completed task can run far above a single chat turn.
- [DeepSeek retires the R1-era API on July 24: your migration options, priced](https://llmhosting.ai/guides/deepseek-r1-retirement-migration): How to migrate off DeepSeek's retiring reasoner API by July 24, priced across V4 Flash, V4 Pro, third-party R1 hosting, and self-hosting the weights.
- [The breakeven math: API tokens vs renting a GPU](https://llmhosting.ai/guides/api-vs-gpu-breakeven-math): The complete formula for deciding whether to pay per token or rent the hardware: the three variables that matter, and the mistakes that flip the answer.
- [Secure, community, marketplace: how to read GPU rental tiers](https://llmhosting.ai/guides/gpu-rental-tiers-explained): The same H100 can cost $1.30 or $3.30 an hour. The difference is the tier: what each one guarantees, and when the cheap one is the right call.
- [H100 vs H200 vs B200 for inference: which should you rent?](https://llmhosting.ai/guides/h100-vs-h200-vs-b200-inference): Three generations of NVIDIA datacenter GPUs are rentable side by side right now. How to choose by workload: memory, bandwidth, or compute.
- [The cheapest way to run DeepSeek R1 in 2026](https://llmhosting.ai/guides/cheapest-way-to-run-deepseek-r1): API, managed endpoints, or renting your own GPU cluster: how to cost out DeepSeek R1's 685B MoE, with the distillation trap and a decision rule to apply today.

## About

- [Methodology](https://llmhosting.ai/about): sources, validation gates, refresh cadence
- [Affiliate disclosure](https://llmhosting.ai/disclosure): some outbound links are referral links; rankings are price-ordered only