Model pricing across providers

AI API price gaps between hosts of the same model

Of the 19 models this directory can compare across API hosts, 2 keep a listed input price gap of 2× or more even without the lowest host. With every host counted, 6 do.

Same developer model, separate services. Hosts may serve different versions, quantizations, context limits or serving tiers of a model. A listing on several providers is not proof of identical weights or behavior. A listed price does not show speed, rate limits or output quality. A host can list a model more than once, at different prices. Each listing counts on its own, under its own name.

DeepInfra has the lowest listing in all 6 models with a gap of 2× or more. As of 2026-11-09, its public model list states a quantization such as fp4 or fp8 for most of its listings here, so its listing may not be the same build another host serves. That is why a model with at least two other hosts also shows its gap without the lowest host, and the models with that figure come first, sorted by it.

Listed token prices by model and host, with the gap between the highest and lowest listed USD price.
Host, listing, price dateInput, per 1M tokensOutput, per 1M tokens
gpt-oss-120bOpenAI · 6 hostsInput gap 3.0× without the lowest host4.1× with every hostOutput gap 3.5×
DeepInfragpt-oss-120bUSD 0.037lowest listedUSD 0.17lowest listed
SiliconFlowgpt-oss-120bUSD 0.05USD 0.45
BasetenGPT OSS 120BUSD 0.1USD 0.5
Amazon Bedrockgpt-oss-120bUSD 0.15highest listedUSD 0.6highest listed
Fireworks AIOpenAI GPT OSS 120BUSD 0.15highest listedUSD 0.60highest listed
Nebius Token Factorygpt-oss-120bUSD 0.15highest listedUSD 0.6highest listed
DeepSeek V4.1 FlashDeepSeek · 4 hostsInput gap 3.0× without the lowest host3.2× with every hostOutput gap 4.3×
DeepInfraDeepSeek V4.1 FlashUSD 0.14lowest listed · sale price; original USD 0.2USD 0.42lowest listed · sale price; original USD 0.6
SiliconFlowDeepSeek-V4.1-FlashUSD 0.15USD 0.6
BasetenDeepSeek V4.1 FlashUSD 0.3USD 1.2
Fireworks AIDeepSeek V4.1 FlashUSD 0.30USD 1.20
Fireworks AIDeepSeek V4.1 Flash (US)USD 0.45highest listedUSD 1.80highest listed
DeepSeek V4 Flash 0731DeepSeek · 3 hostsInput gap 1.7× without the lowest host3.7× with every hostOutput gap 3.7×
DeepInfraDeepSeek V4 Flash 0731USD 0.06lowest listedUSD 0.18lowest listed
BasetenDeepSeek-V4-Flash-0731USD 0.13USD 0.26
SiliconFlowDeepSeek-V4-Flash-0731USD 0.22highest listedUSD 0.66highest listed
Kimi K3Moonshot AI · 7 hostsInput gap 1.6× without the lowest host1.7× with every hostOutput gap 1.7×
SiliconFlowKimi-K3USD 2.7lowest listedUSD 13.5lowest listed
DeepInfraKimi K3USD 2.85USD 14.25
BasetenKimi K3USD 3USD 15
Fireworks AIKimi K3USD 3.00USD 15.00
Moonshot AIKimi K3USD 3.00USD 15.00
Nebius Token FactoryKimi-K3USD 3USD 15
Amazon BedrockKimi K3USD 3.3USD 16.5
Fireworks AIKimi K3 (US)USD 4.50highest listedUSD 22.50highest listed
GLM-5.3Z.ai · 6 hostsInput gap 1.5× without the lowest host3.7× with every hostOutput gap 2.6×
DeepInfraGLM 5.3USD 0.5625lowest listed · sale price; original USD 0.9USD 2.5lowest listed · sale price; original USD 4
BasetenGLM-5.3USD 1.4USD 4.4
Fireworks AIGLM 5.3USD 1.40USD 4.40
Nebius Token FactoryGLM-5.3USD 1.4USD 4.4
SiliconFlowGLM-5.3USD 1.4USD 4.4
Z.aiGLM-5.3USD 1.4USD 4.4
Fireworks AIGLM 5.3 (US)USD 2.10highest listedUSD 6.60highest listed
GLM-5.3 FlashZ.ai · 6 hostsInput gap 1.5× without the lowest host1.5× with every hostOutput gap 1.5×
BasetenGLM-5.3-FlashUSD 0.15lowest listedUSD 0.5lowest listed
DeepInfraGLM 5.3 FlashUSD 0.15lowest listedUSD 0.5lowest listed
Fireworks AIGLM 5.3 FlashUSD 0.15lowest listedUSD 0.50lowest listed
Nebius Token FactoryGLM-5.3-FlashUSD 0.15lowest listedUSD 0.5lowest listed
SiliconFlowGLM-5.3-FlashUSD 0.15lowest listedUSD 0.5lowest listed
Z.aiGLM-5.3-FlashUSD 0.15lowest listedUSD 0.50lowest listed
Fireworks AIGLM 5.3 Flash (US)USD 0.225highest listedUSD 0.75highest listed
GLM-5.2Z.ai · 5 hostsInput gap 1.1× without the lowest host2.5× with every hostOutput gap 2.4×
DeepInfraGLM 5.2USD 0.5625lowest listed · sale price; original USD 0.75USD 1.8lowest listed · sale price; original USD 2.4
SiliconFlowGLM-5.2USD 1.302USD 4.092
BasetenGLM-5.2USD 1.4highest listedUSD 4.4highest listed
Nebius Token FactoryGLM-5.2USD 1.4highest listedUSD 4.4highest listed
Z.aiGLM-5.2USD 1.4highest listedUSD 4.4highest listed
MiniMax M3MiniMax · 5 hostsInput gap 1.0× without the lowest host1.1× with every hostOutput gap 1.1×
DeepInfraMiniMax M3USD 0.28lowest listedUSD 1.1lowest listed
Fireworks AIMiniMax M3USD 0.30highest listedUSD 1.20highest listed
MiniMaxMiniMax-M3USD 0.30highest listed · ≤512K contextUSD 1.20highest listed · ≤512K context
Nebius Token FactoryMiniMax-M3USD 0.3highest listedUSD 1.2highest listed
SiliconFlowMiniMax-M3USD 0.3highest listedUSD 1.2highest listed
GLM-5Z.ai · 3 hostsInput gap 1.0× without the lowest host1.1× with every hostOutput gap 1.3×
SiliconFlowGLM-5USD 0.95lowest listedUSD 2.55lowest listed
Amazon BedrockGLM 5USD 1highest listedUSD 3.2highest listed
Novita AIGLM 5USD 1highest listed · Serverless tier · availability unverifiedUSD 3.2highest listed · Serverless tier · availability unverified
DeepSeek V4 Pro 0813DeepSeek · 5 hostsInput gap 1.0× without the lowest host1.0× with every hostOutput gap 1.5×
DeepInfraDeepSeek V4 Pro 0813USD 1.3lowest listedUSD 2.6lowest listed
BasetenDeepSeek V4 Pro 0813USD 1.32highest listedUSD 3.96highest listed
DeepSeekDeepSeek V4 Pro 0813USD 1.32highest listed · peakUSD 3.96highest listed · peak
Nebius Token FactoryDeepSeek-V4-Pro-0813USD 1.32highest listedUSD 3.96highest listed
SiliconFlowDeepSeek-V4-Pro-0813USD 1.32highest listedUSD 3.96highest listed
DeepSeek V4.1 Flash 1017DeepSeek · 2 hostsInput gap 1.0× without the lowest host1.0× with every hostOutput gap 1.0×
BasetenDeepSeek V4.1 Flash 1017USD 0.3USD 1.2
Nebius Token FactoryDeepSeek V4.1 Flash 1017USD 0.3USD 1.2
MiMo V2.6 FlashXiaomi MiMo · 2 hostsInput gap 1.0× without the lowest host1.0× with every hostOutput gap 1.0×
DeepInfraMiMo V2.6 FlashUSD 0.14USD 0.28
Xiaomi MiMomimo-v2.6-flashUSD 0.14USD 0.28
Qwen3.8-MaxQwen (Alibaba Cloud) · 2 hostsInput gap 1.0× without the lowest host1.0× with every hostOutput gap 1.0×
Alibaba Cloud Model StudioQwen3.8-MaxUSD 2≤1M tokensUSD 6≤1M tokens
Fireworks AIQwen 3.8 MaxUSD 2.00USD 6.00
Qwen3.6-35B-A3BQwen (Alibaba Cloud) · 2 hostsInput gap 2.4×Output gap 1.9×
DeepInfraQwen3.6 35B A3BUSD 0.1lowest listedUSD 0.95lowest listed
SiliconFlowQwen3.6-35B-A3BUSD 0.24highest listedUSD 1.8highest listed
gpt-oss-20bOpenAI · 2 hostsInput gap 1.8×Output gap 1.7×
SiliconFlowgpt-oss-20bUSD 0.04lowest listedUSD 0.18lowest listed
Amazon Bedrockgpt-oss-20bUSD 0.07highest listedUSD 0.3highest listed
Kimi K2.5Moonshot AI · 2 hostsInput gap 1.3×Output gap 1.3×
SiliconFlowKimi-K2.5USD 0.45lowest listedUSD 2.25lowest listed
Amazon BedrockKimi K2.5USD 0.6highest listedUSD 3highest listed
MiMo V2.6 ProXiaomi MiMo · 2 hostsInput gap 1.1×Output gap 1.1×
DeepInfraMiMo V2.6 ProUSD 0.387lowest listed · sale price; original USD 0.43USD 0.783lowest listed · sale price; original USD 0.87
Xiaomi MiMomimo-v2.6-proUSD 0.435highest listedUSD 0.87highest listed
Grok 4.7xAI · 2 hostsInput gap 1.1×Output gap 1.1×
xAIgrok-4.7USD 2.00lowest listed · <200K tokensUSD 6.00lowest listed · <200K tokens
Amazon BedrockGrok 4.7USD 2.2highest listedUSD 6.6highest listed
Kimi K2.6Moonshot AI · 2 hostsInput gap 1.0×Output gap 1.0×
DeepInfraKimi K2.6USD 0.75lowest listedUSD 3.5highest listed
SiliconFlowKimi-K2.6USD 0.77highest listedUSD 3.4lowest listed
How the gaps are calculated

A gap is the highest listed price divided by the lowest listed price for the same model, using each listing's headline rate under the directory's shared rule. Input and output are calculated separately. A headline rate can be conditional, such as a peak-hours tier or a context limit, which the row names. These prices were observed from 2026-10-26 to 2026-11-09.

Only USD prices are compared, and no currency conversion is performed. A sale price counts as listed and is marked. A stale price, an ended price or a free listing is shown but not compared, and a gap needs at least two hosts.

"Without the lowest host" removes every listing from the one host that sets the lowest price, then calculates the gap again from at least two remaining hosts. When hosts share the lowest price, no single host sets it, so the gap stands. Listings are matched by reviewed evidence, as on each model's comparison page. See the methodology.