IBM
2 models built
IBM is a century-old enterprise technology company (founded 1911) that entered the open-weight LLM space with its Granite family, built for business use rather than chatbot benchmarks. On OpenKey you get two Granite models: Granite 4.0 Micro and Granite 4.1 8B, both text-to-text, both with 131K-token context windows. IBM's strategy with Granite is small, cheap, deployable models rather than a race for the largest parameter count — these are the kind of models you'd run for document processing, internal tools, or high-volume tasks where cost per call matters more than chasing a leaderboard. Neither model is free, but both are priced low enough to run at scale.
Lineup
With only two models, the choice comes down to cost versus capability. Granite 4.0 Micro is the cheaper option at $0.017/M input and $0.112/M output — use it for high-volume, straightforward text tasks where every fraction of a cent adds up. Granite 4.1 8B costs more on input ($0.05/M) but less on output ($0.10/M) and is the newer of the two (April 2026 vs October 2025), so lean on it if you want IBM's more current model or have output-heavy workloads. Both share the same 131K-token context length, so context size isn't a differentiator here — pick based on your input/output ratio and how recent you want the model to be.
- Granite 4.1 8BApr 30, 2026 131K $0.051 $0.103 ctx 131K · in $0.051 · out $0.103
- Granite 4.0 MicroOct 20, 2025 131K $0.018 $0.115 ctx 131K · in $0.018 · out $0.115
Prices per 1M tokens, flat 3% fee included.
Questions
- What does IBM Granite cost on OpenKey?
- Granite 4.0 Micro runs $0.017/M input and $0.112/M output on the provider list; on OpenKey that's $0.0175/M input and $0.1154/M output (price × 1.03). Granite 4.1 8B is $0.05/M input and $0.10/M output from IBM, or $0.0515/M and $0.103/M on OpenKey. Both are priced for high-volume use.
- Which Granite model should I use for cheap, high-volume tasks?
- Granite 4.0 Micro. At $0.017/M input it's the cheapest input price of the two, making it the better fit if your workload is input-heavy — think bulk document processing or classification. Granite 4.1 8B costs more on input but less on output, so it's better suited to output-heavy jobs.
- Are any IBM Granite models free on OpenKey?
- No. Neither Granite 4.0 Micro nor Granite 4.1 8B is free — IBM has zero free models in its two-model lineup on OpenKey, out of the platform's 25 free models overall. Both are still low-cost compared to most other labs' offerings.