Models / gemma-4-31b

Gemma 4 31B

Google's model, with open weights. Forcebench has run it in 1 configuration on the Local board, plus 1 on the lite subset. API is models reached through a hosted service; Local, models served on Forcebench’s own hardware.

On the Local board

Models served on Forcebench’s own hardware. Gemma 4 31B is ranked by its one configuration: the sixth-highest score of 9 models, not clearly different from 5 other models. Scores are within 3 attempts (c@3) with 95% intervals; each configuration is compared with the highest-scoring one by whether their 95% intervals overlap. Every configuration on the board →

43%
within 3 attempts (c@3), 95% CI 38–48 · thinking on · sixth-highest score of 9 models, not clearly different from 5 other models
2.8k
output tokens per task · median 2.3k, p90 5.3k
42 s
median per task, every attempt counted, on 1× NVIDIA GeForce RTX 5090 (32 GB) · 90th percentile 240 s
0%
no answer within the 32,768-token budget
Every configuration of Gemma 4 31B on the Local board
ConfigurationRank all configsWithin 3 attempts c@3 · 95% CITokens mean · median · p90No answerLatency median per task, every attempt counted
QAT W4A16 · vLLM · thinking on · 1× NVIDIA GeForce RTX 5090 (32 GB)highest score13
43%95% CI 38–48
2.8k2.3k · 5.3k0%42 s

Latency depends on the hardware: compare it only between configurations on the same hardware.

Suites, at its highest-scoring configuration
thinking on · QAT W4A16 · vLLM · 1× NVIDIA GeForce RTX 5090 (32 GB): within 3 attempts (c@3) on each suite with 95% intervals, against the median of the Local board's models, each at its highest-scoring configuration.
  • 95% interval
  • Local board median
Apex
70%, 95% interval 48–85%
Salesforce APIs
50%, 95% interval 29–71%
CI/CD
59%, 95% interval 36–78%
Salesforce CLI
15%, 95% interval 5–36%
Salesforce docs
0%, 95% interval 0–18%
fflib Enterprise Patterns
17%, 95% interval 6–39%
Flow
11%, 95% interval 3–33%
Governor limits & pushback
89%, 95% interval 69–97%
Lightning Web Components
89%, 95% interval 67–97%
Nonprofit Success Pack
61%, 95% interval 39–80%
Packaging
17%, 95% interval 6–39%
Permissions & access
47%, 95% interval 25–70%
Scratch org definitions
6%, 95% interval 1–26%
SOQL
75%, 95% interval 53–89%
Trigger Actions Framework
39%, 95% interval 20–61%
    Show the numbers
    Per-suite scores of Gemma 4 31B's highest-scoring configuration on the Local board, with the board's median
    Suitec@3 95% CILocal medianTasks
    Apex
    70%95% CI 48–85
    		70%20
    Salesforce APIs
    50%95% CI 29–71
    		44%18
    CI/CD
    59%95% CI 36–78
    		53%17
    Salesforce CLI
    15%95% CI 5–36
    		45%20
    Salesforce docs
    0%95% CI 0–18
    		12%17
    fflib Enterprise Patterns
    17%95% CI 6–39
    		17%18
    Flow
    11%95% CI 3–33
    		17%18
    Governor limits & pushback
    89%95% CI 69–97
    		84%19
    Lightning Web Components
    89%95% CI 67–97
    		89%18
    Nonprofit Success Pack
    61%95% CI 39–80
    		61%18
    Packaging
    17%95% CI 6–39
    		39%18
    Permissions & access
    47%95% CI 25–70
    		60%15
    Scratch org definitions
    6%95% CI 1–26
    		17%18
    SOQL
    75%95% CI 53–89
    		90%20
    Trigger Actions Framework
    39%95% CI 20–61
    		44%18

    In the studies

    Private evaluations

    Want this for your team?

    Azul Labs runs Forcebench privately for two kinds of Salesforce team.

    Choosing a model to run yourself?

    A private benchmark of your shortlist

    Open-weight models at the quantisations and reasoning efforts you could run on infrastructure you control, tested on your kind of Salesforce work, with a written recommendation.

    • Private tasks written from your work only ever run on local models; they are never sent to an API.
    • Re-testing after any change of model, quantisation or engine is available.
    • Typical turnaround is 2–3 weeks; large models take about a day of machine time per configuration.
    Get a private benchmarkHow it works

    Building or buying a Salesforce AI tool?

    A private evaluation of your tool

    Your tool on held-out tasks it has never seen, against API baselines (models reached through a hosted service) and Local ones (served on Forcebench’s own hardware), with a written report on where it holds up and where it fails.

    • It can start as a smaller pilot, about 30 held-out tasks with your tool against two baselines, before a full evaluation.
    • Repeat evaluations after your releases are available.
    • Typical turnaround is 2–3 weeks.
    Ask about an evaluationHow it works

    Fixed fee, quoted up front. The quote depends on the number of tasks and the number of models, tools or configurations tested.

    Or email leo@azl.au · azl.au