Granite 4 H Medium vs GPT-5.6 Luna
Side-by-side comparison of API pricing, specs, benchmarks, and capabilities
| Specification |
by IBM
|
by OpenAI
|
|---|---|---|
| Overview | ||
| Status | Current mid tier Open weights | Current mid tier |
| Released | Oct 19, 2025 | Jul 9, 2026 |
| Pricing per million tokens | ||
| Input | $0.150/Mtok | $0.200/Mtok |
| Output | $0.600/Mtok | $1.20/Mtok |
| Blended avg | $0.375/Mtok | $0.700/Mtok |
| Cached input | — | $0.020/Mtok |
| Specifications | ||
| Context window | 128K tokens | 1M tokens |
| Parameters | Proprietary | Proprietary |
| Speed (TPS) | — | — |
| Modalities | ||
| Input |
text
|
textimage
|
| Benchmarks sources: IBM model card / Vellum LLM Leaderboard | ||
| GPQA Diamond | 35 vendor | 92.3 |
| SWE-Bench Verified | 15 vendor | 93 |
| HumanEval | 72 vendor | 93 |
| Avg benchmark score | — | 90.7 |
| Perf / dollar | — | 25.9 |
| Terminal-Bench 2.1 | — | 84.3 |
| Humanity's Last Exam | 7 vendor | — |
| ARC-AGI 2 | 5 vendor | — |
| AIME 2025 | 28 vendor | — |
| MMMLU | 68 vendor | — |
| BFCL | 48 vendor | — |
| MATH 500 | 50 vendor | — |
| Independent composite scores each on its own scale — not part of the average above | ||
| AA Intelligence Index | — | 51.2 |
| AA Agentic Index | — | 45.6 |
| AA-Omniscience Index (−100–100) | — | -11.2 |
| GDPval-AA v2 | — | 1581.9 ELO |
| Providers | ||
| Available from |
IBM — $0.150/$0.600/Mtok
|
OpenAI — $0.200/$1.20/Mtok
|
Cost at scale
1M tokens · 50/50 input/output| Volume | Granite 4 H Medium | GPT-5.6 Luna | Savings |
|---|---|---|---|
| 1M tokens | $0.38 | $0.7 | $0.32 (45.7%) |
| 10M tokens | $3.75 | $7 | $3.25 (46.4%) |
| 100M tokens | $37.5 | $70 | $32.5 (46.4%) |
| 1000M tokens | $375 | $700 | $325 (46.4%) |
When to pick which
distilled from the pricing and spec data above- cost dominates: $0.375/Mtok blended vs $0.700 — 46.4% less on the same 50/50 token mix
- you want open weights — self-host it, fine-tune it, or exit the API entirely; GPT-5.6 Luna is closed
- your workload re-reads context (agents, RAG, long chats): cached input costs $0.020/Mtok — 90% off its list input price, a discount Granite 4 H Medium doesn't offer
- you need the longer context: 1M tokens vs 128K (8.2×)
Price movement
recorded changes since we began tracking each model · last checked Aug 1, 2026| Model | Changes | Latest move | Old → new $/M |
|---|---|---|---|
| Granite 4 H Medium | 0 | — no move | Stable since we began tracking — no change recorded |
| GPT-5.6 Luna | 1 | price cut on Aug 1, 2026 | $1/$6 → $0.2/$1.2/Mtok |
The cheaper option here — Granite 4 H Medium costs $0.375/Mtok blended on IBM.
Summary
Granite 4 H Medium by IBM costs $0.150/Mtok input and $0.600/Mtok output, with a 128K-token context window. It supports text input.
GPT-5.6 Luna by OpenAI costs $0.200/Mtok input and $1.20/Mtok output, with a 1M-token context window. It supports text, image input.
On a blended cost basis, Granite 4 H Medium is 46.4% cheaper than GPT-5.6 Luna.
The two aren't directly comparable on average benchmark score: GPT-5.6 Luna has published per-benchmark results, while Granite 4 H Medium does not yet.
Note: Pricing is per million tokens. Actual costs vary with usage patterns, prompt caching, and batch discounts. Always verify against official provider pricing pages.