Pricing Watch / Cheapest by job
Cheapest sourced SKU by agent job (74 jobs)
Each page compares candidates that share a single native billing unit. Prices are never blended or converted across units, and the cheapest SKU is always distinct from the editorial Depot Pick.
Frontier general chat & reasoning
usd_per_million_input_tokensLowest: Z.AI (Zhipu GLM) ($0.04 · GLM-4.6V-FlashX)
Coding agent / code generation
usd_per_million_input_tokensLowest: Z.AI (Zhipu GLM) ($0.04 · GLM-4.6V-FlashX)
Long-context agent workflows
usd_per_million_input_tokensLowest: Anthropic ($0.1 · Haiku 5.5 (prompts up to 100K tokens))
Tool-calling production agents
usd_per_million_input_tokensLowest: Anthropic ($0.1 · Haiku 5.5 (prompts up to 100K tokens))
High-volume cheap LLM inference
usd_per_million_input_tokensLowest: Cloudflare Workers AI ($0.017 · @cf/ibm-granite/granite-4.0-h-micro)
Hosted open-weight model inference
usd_per_million_input_tokensLowest: Chutes ($0.039 · Qwen3.8-27B (Qwen/Qwen3.8-27B-TEE))
Ultra-low-latency token streaming
usd_per_million_input_tokensLowest: Groq ($0.075 · openai/gpt-oss-20b)
Multi-provider model routing
usd_per_million_input_tokensLowest: Amazon Bedrock ($0.04 · Gemma 3 4B)
Enterprise-governed LLM access
usd_per_million_input_tokensLowest: Amazon Bedrock ($0.04 · Gemma 3 4B)
EU-friendly / sovereign LLM
usd_per_million_input_tokensLowest: Anthropic ($0.1 · Haiku 5.5 (prompts up to 100K tokens))
Enterprise RAG stack
usd_per_million_input_tokensLowest: Amazon Bedrock ($0.04 · Gemma 3 4B)
Multimodal understanding (vision+)
usd_per_million_input_tokensLowest: Z.AI (Zhipu GLM) ($0.04 · GLM-4.6V-FlashX)
Serverless GPU inference/jobs
usd_per_gpu_hourLowest: RunPod ($0.34 · RTX 4090 Pod (Community Cloud))
Cheapest on-demand GPUs
usd_per_gpu_hourLowest: SaladCloud ($0.09 · RTX 3090 24GB (Lowest priority, batch))
Predictable dedicated GPU capacity
usd_per_gpu_hourLowest: RunPod ($0.34 · RTX 4090 Pod (Community Cloud))
Multi-node GPU training
usd_per_gpu_hourLowest: Crusoe ($1.5 · NVIDIA L40S 48GB)
Enterprise GPU cloud
usd_per_gpu_hourLowest: Crusoe ($1.5 · NVIDIA L40S 48GB)
Batch / interruptible GPU jobs
usd_per_gpu_hourLowest: SaladCloud ($0.09 · RTX 3090 24GB (Lowest priority, batch))
Fine-tune open models
usd_per_million_train_tokensLowest: Fireworks AI ($0.5 · Managed training LoRA SFT up to 16B)
Cloud browser for agents
usd_per_browser_hourLowest: Browser Use Cloud ($0.02 · Browsers)
Self-host browser automation
usd_per_browser_hourLowest: Steel.dev ($0.08 · Scale browser hours)
SERP / search results API
usd_per_1k_searchesLowest: Serper ($1 · Google SERP API, Starter top-up)
Autonomous web-acting agent
usd_per_browser_hourLowest: Browser Use Cloud ($0.02 · Browsers)
Persistent agent dev environments
usd_per_vcpu_hourLowest: Daytona ($0.0504 · vCPU)
Transactional email
usd_per_1k_emailsLowest: Amazon SES ($0.1 · Outbound email (à la carte) transactional)
Marketing / blast email
usd_per_emailLowest: SendGrid (Twilio) ($0.0023 · Marketing Campaigns Basic 100K extra email or contact)
Programmable SMS
usd_per_sms_segmentLowest: Twilio ($0.0083 · US SMS outbound, long code)
Programmable voice / phone agents
usd_per_minuteLowest: Plivo ($0.0115 · US local calls, to make calls (outbound PSTN))
Programmatic image generation
usd_per_imageLowest: MiniMax ($0.0035 · image-01)
Programmatic video generation
usd_per_videoLowest: Luma AI ($0.15 · Ray3.2 text/image-to-video, 5s, 540p)
Text-to-speech
usd_per_1k_charactersLowest: Deepgram ($0.03 · Aura-2)
Speech-to-text / transcription
usd_per_minuteLowest: Cloudflare Workers AI ($0.0005 · @cf/openai/whisper audio minute)
Voice agents (talking products)
usd_per_minuteLowest: Vapi ($0.05 · Vapi hosting)
Web search tool for agents
usd_per_1k_searchesLowest: Serper ($1 · Google SERP API, Starter top-up)
Neural / semantic web search
usd_per_1k_searchesLowest: Brave Search API ($5 · Search API)
Search-augmented generation
usd_per_searchLowest: Exa ($0.005 · Agent search tool call)
Privacy-oriented search API
usd_per_1k_searchesLowest: Brave Search API ($5 · Search API)
Enterprise web search API
usd_per_1k_searchesLowest: Brave Search API ($5 · Search API)
Object storage for agent artifacts
usd_per_gb_monthLowest: Cloudflare R2 ($0.015 · R2 Standard storage)
Object storage without egress pain
usd_per_gb_monthLowest: Cloudflare R2 ($0.015 · R2 Standard storage)
Ultra-budget open-model inference
usd_per_million_input_tokensLowest: Chutes ($0.039 · Qwen3.8-27B (Qwen/Qwen3.8-27B-TEE))
Edge-colocated LLM inference
usd_per_million_input_tokensLowest: Cloudflare Workers AI ($0.017 · @cf/ibm-granite/granite-4.0-h-micro)
LLM on existing lakehouse data
usd_per_million_input_tokensLowest: Amazon Bedrock ($0.04 · Gemma 3 4B)
Self-hosted optimized model serving
usd_per_million_input_tokensLowest: Baseten ($0.1 · GPT OSS 120B)
Neocloud reserved / list-price GPUs
usd_per_gpu_hourLowest: Massed Compute ($0.97 · L40S (48GB) x1)
Ultra-cheap interruptible GPUs
usd_per_gpu_hourLowest: SaladCloud ($0.09 · RTX 3090 24GB (Lowest priority, batch))
Multi-cloud GPU shopping / routing
usd_per_gpu_hourLowest: RunPod ($0.34 · RTX 4090 Pod (Community Cloud))
Computer-use / stealth cloud browsers
usd_per_browser_hourLowest: Browser Use Cloud ($0.02 · Browsers)
Managed browser-agent loop
usd_per_browser_hourLowest: Browser Use Cloud ($0.02 · Browsers)
Ultra-low-latency text-to-speech
usd_per_1k_charactersLowest: Deepgram ($0.03 · Aura-2)
Production speech-to-text
usd_per_minuteLowest: Deepgram ($0.0043 · Nova-3 Monolingual pre-recorded speech-to-text)
Full voice-agent platform (STT+TTS+orche…
usd_per_minuteLowest: Vapi ($0.05 · Vapi hosting)
Content moderation (I/O safety classify)
usd_per_1k_requestsLowest: OpenAI Moderation ($0 · omni-moderation-latest (moderation endpoint))
Programmable LLM input/output guardrails
usd_per_1k_requestsLowest: OpenAI Moderation ($0 · omni-moderation-latest (moderation endpoint))
Multilingual / code-switching speech-to-text
usd_per_minuteLowest: Deepgram ($0.0043 · Nova-3 Monolingual pre-recorded speech-to-text)
Parse documents into LLM-ready chunks
usd_per_1k_pagesLowest: Reducto ($10 · r-1 Parse API)
Rerank retrieved chunks for RAG
usd_per_million_tokensLowest: Jina AI ($0.02 · Jina Reader / Embeddings / Reranker Token Pack ($0.02 / 1M tokens))
Reliable outbound webhooks for agent products
usd_per_messageLowest: Svix ($0.0001 · Additional messages)
TEE inference
usd_per_million_input_tokensLowest: Chutes ($0.039 · Qwen3.8-27B (Qwen/Qwen3.8-27B-TEE))
Batch embeddings
usd_per_million_tokensLowest: Voyage AI ($0.02 · voyage-4-lite)
Prompt caching
usd_per_million_input_tokensLowest: Amazon Bedrock ($0.04 · Gemma 3 4B)
Agent telephony
usd_per_minuteLowest: Plivo ($0.0115 · US local calls, to make calls (outbound PSTN))
Layout OCR
usd_per_1k_pagesLowest: Reducto ($10 · r-1 Parse API)
PR review agents
usd_per_million_input_tokensLowest: Anthropic ($0.1 · Haiku 5.5 (prompts up to 100K tokens))
LoRA / distill host
usd_per_million_train_tokensLowest: Fireworks AI ($0.5 · Managed training LoRA SFT up to 16B)
Agent DLQ
usd_per_100k_messagesLowest: Upstash QStash ($1 · Pay as You Go messages)
Low-carbon GPU
usd_per_gpu_hourLowest: SaladCloud ($0.09 · RTX 3090 24GB (Lowest priority, batch))
HIPAA LLM
usd_per_million_input_tokensLowest: Amazon Bedrock ($0.04 · Gemma 3 4B)
Air-gapped serve
usd_per_million_input_tokensLowest: Amazon Bedrock ($0.04 · Gemma 3 4B)
Edge small models
usd_per_million_input_tokensLowest: Cloudflare Workers AI ($0.017 · @cf/ibm-granite/granite-4.0-h-micro)
Inbound webhooks
usd_per_messageLowest: Svix ($0.0001 · Additional messages)
Deep research
usd_per_searchLowest: Exa ($0.005 · Agent search tool call)
Text-in-image
usd_per_imageLowest: Black Forest Labs ($0.009 · FLUX 3 Image Klein (1K tier))
Budget batch open models
usd_per_gpu_hourLowest: SaladCloud ($0.09 · RTX 3090 24GB (Lowest priority, batch))