Model
Model explorer

Granite 34B Code Instruct

OPEN
IBM · Granite Code family · released May 6, 2024

Largest Granite Code model, depth-upscaled from Granite-20B-Code; now deprecated by IBM in favor of newer Granite releases.

ReasoningCodingVisionFunction callingTool useAgentic
424.5
Elo · unrated
Parameters
34B
Active params
34B (dense)
Context
8K tokens
Architecture
Dense decoder-only transformer (depth-upscaled from Granite-20B-Code)
License
Apache 2.0
Languages
—
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
HumanEvalCoding54.8%#112
best: Claude Opus 4.5 · 99.4%
Run it locally
VRAM @ Q4
21.4 GB
VRAM @ FP16
68 GB
Fits on (Q4)
RTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Throughput data unavailable.
Quantizations
GGUF
Fine-tune it
Permissive
QLoRA23.4 GB1× RTX 3090 24GB
LoRA72.7 GB1× A100 80GB
Full fine-tune546.0 GB4× H200 141GB
QLoRA SFT on ~10k samples ≈ $17.51 (1× RTX 3090 24GB)
Plan a fine-tune →
Granite Code family
Elo progression across releases
Granite 34B Code InstructMay 2024424.5Granite 8B Code InstructMay 2024296.4
Resources
HFibm-granite/granite-34b-code-instruct-8kGHibm-granite/granite-code-modelsDOCSibm.com/granite/docs/models/code
API price weights · each benchmark row carries its own source badge (see methodology)