Skip to content

status.openkey.ai is live — independent, worldwide uptime proof See it now ->

openkey

Microsoft

2 models built

Microsoft Corporation, founded 1975, is OpenAI's largest commercial partner and builds the Phi family of small language models alongside its Copilot products. On OpenKey the lineup is small: just two models. Phi 4, released January 2025, is the current small-model flagship, built for efficiency over raw scale. WizardLM-2 8x22B, dated April 2024, is a larger mixture-of-experts model with a longer context window. Neither is free. This isn't a broad frontier-model catalog — it's a narrow set reflecting Microsoft's small-model research plus one older community-adjacent release, rather than a full commercial API business the way OpenAI or Anthropic run theirs.

Lineup

With only two models, the choice comes down to context length and cost. Phi 4 is cheap and compact: $0.0721/1M input tokens and $0.1442/1M output tokens on OpenKey (provider price $0.07 and $0.14, plus the 3% fee), with a 16,384-token context window — fine for short-to-medium tasks where cost matters. WizardLM-2 8x22B costs more at $0.6386/1M tokens for both input and output ($0.62 x 1.03), but gives you a 65,536-token context window, useful if you need to feed in longer documents or conversations. There's no free option here, so pick based on whether you need Phi 4's low cost or WizardLM-2's longer context.

  • Phi 4 ctx 16K · in $0.072 · out $0.144
  • WizardLM-2 8x22B ctx 66K · in $0.639 · out $0.639

Prices per 1M tokens, flat 3% fee included.

Questions

What does Microsoft charge on OpenKey?
Two models, two price points. Phi 4 runs $0.0721/1M input and $0.1442/1M output tokens. WizardLM-2 8x22B runs $0.6386/1M tokens for both input and output. Both figures are the provider's list price plus OpenKey's flat 3% fee — no other markup.
Which Microsoft model should I use?
Phi 4 is the newer, cheaper choice at $0.0721/1M input and $0.1442/1M output, good for cost-sensitive work within a 16,384-token context. WizardLM-2 8x22B costs more ($0.6386/1M both directions) but supports a 65,536-token context, so reach for it when the task needs more room.
Are any Microsoft models free on OpenKey?
No. Microsoft has zero free models in the OpenKey catalog — both Phi 4 and WizardLM-2 8x22B are paid, priced at provider list plus the flat 3% fee. If you need a free model, look at other labs in the 25-model free tier.