Gemma 4 31B vs Granite 4.2 8B

Practical read: Gemma 4 31B reads two times more context. Granite 4.2 8B costs about half as much. One API key runs both.

No card required. Start with 1 million free tokens.

Side by side:

Google: Gemma 4 31B compared with IBM: Granite 4.2 8B
Attribute	Gemma 4 31B `google/gemma-4-31b-it`	Granite 4.2 8B `ibm-granite/granite-4.2-8b`
Price, 1M in + out	$0.46 Weights & Biases, cheapest route	$0.22 Weights & Biases, cheapest route
Best for	Long documents	Low cost
Input / output, 1M	$0.11 / $0.36	$0.063 / $0.16
Cached input	$0.060927/1M to $0.26375/1M	$0.015825/1M
Context window	262K 262,144 tokens	131K 131,072 tokens
First token, p50	0.78 s via Tinfoil	Measuring
Throughput	250 tok/s via FriendliAI	Measuring
AI IQ	IQ 101 #120 on AI IQ	IQ 94 #151 on AI IQ
Provider routes	14 with failover	2 with failover
Privacy	Confidential E2EE and zero retention route	Varies by provider
Recent route uptime range	100.00%	not enough data
Publisher	Google AI Studio	IBM
Modes	chat	chat
More detail	Overview Pricing Providers	Overview Pricing Providers

Price adds one million input and one million output tokens on the cheapest route. Prices and speed change: see the monthly reports and run a small eval before you choose a default.

Test both with your agent

Paste this into Claude Code, Cursor or any coding agent.

Compare Gemma 4 31B and Granite 4.2 8B in this project through TrustedRouter.

1. Read https://trustedrouter.com/docs/llms.txt
2. Point the OpenAI-compatible client at
   https://api.trustedrouter.com/v1 and read the key from
   TRUSTEDROUTER_API_KEY.
3. Run our eval prompts on google/gemma-4-31b-it, then on ibm-granite/granite-4.2-8b.
4. Show me cost, latency and quality side by side before changing the default.

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.trustedrouter.com/v1",
    api_key=os.environ["TRUSTEDROUTER_API_KEY"],
)
for model in ("google/gemma-4-31b-it", "ibm-granite/granite-4.2-8b"):
    response = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": "What is the capital of France?"}],
    )
    print(model, response.choices[0].message.content)

for model in google/gemma-4-31b-it ibm-granite/granite-4.2-8b; do
  curl https://api.trustedrouter.com/v1/chat/completions \
    -H "Authorization: Bearer $TRUSTEDROUTER_API_KEY" \
    -H "Content-Type: application/json" \
    -d "{\"model\": \"$model\", \"messages\": [{\"role\": \"user\", \"content\": \"What is the capital of France?\"}]}"
done

Your agent will need an API key.

Questions

Which should I use, Google: Gemma 4 31B or IBM: Granite 4.2 8B?

Google: Gemma 4 31B has more Credits provider routes and the larger context window. IBM: Granite 4.2 8B has the lower published input-plus-output rate. There is not enough recent probe data to compare speed.

Is Google: Gemma 4 31B or IBM: Granite 4.2 8B cheaper?

The current cheapest TrustedRouter route is $0.4642/1M for Google: Gemma 4 31B and $0.22155/1M for IBM: Granite 4.2 8B. The comparison uses current catalog prices and updates as provider pricing changes.

Is Google: Gemma 4 31B or IBM: Granite 4.2 8B faster?

Current measured p50 time to first token is 781 ms for Google: Gemma 4 31B and not enough data for IBM: Granite 4.2 8B. These are routed probe measurements, not vendor-advertised speeds, and update as new samples arrive.

Can I test Google: Gemma 4 31B and IBM: Granite 4.2 8B with the same API?

Yes. Use the same OpenAI-compatible TrustedRouter base URL and API key, then change only the model id between google/gemma-4-31b-it and ibm-granite/granite-4.2-8b. This makes side-by-side evals possible without maintaining two provider integrations.

More comparisons

All comparisons

Try Gemma 4 31B and Granite 4.2 8B with one key.

TrustedRouter is useful when you know the model you want, but still need provider rollover, budget limits, usage records, and a prompt path you can verify.

No card required. Start with 1 million free tokens.