Standard Compute
Model comparison 2026

Beam 1.0 vs Devstral 2

Beam 1.0
Reflection · Balanced · Open weights
$1.60/1M · 1.0M ctx
Devstral 2
Mistral · Balanced · Open weights
$0.88/1M · 262K ctx

Community vote

Pick a winner in each dimension. You can change your vote any time.

Output Quality

Correctness and depth of what it produces

Agentic Ability

Tool calls, instruction-following, and multi-step tasks

Speed

Tokens per second and time-to-first-token

Value for Money

How much capability you get per dollar

Reliability

Consistent results: fewer refusals, loops and format breaks

Specs and pricing side by side

Spec
Beam 1.0
Devstral 2
Maker
Reflection
Mistral
Blended price / 1M
$1.60
$0.88
Input / output
$1.00 in · $3.00 out / 1M tokens
$0.40 in · $2.00 out / 1M tokens
Context window
1.0M
262K
Open weights
Yes
Yes
Tool use
Yes
Yes
Reasoning
Yes
No
Output Quality
7.0
3.3
Agentic Ability
5.0
4.4
Speed
7.0
8.2
Value for Money
5.7
5.7
Reliability
8.0
8.4

Pricing and capabilities synced from the OpenRouter catalogue. Scores are editorial (0–10).

Verdict: Beam 1.0 or Devstral 2?

Updated 2026-10-17

Choose Beam 1.0 if you want trying a permissively licensed open-weight coding model through OpenRouter, or self-hosting it where code must stay on your own hardware. Choose Devstral 2 if you want coding agents that need a Mistral open-weight model.

Editorially it's close: each model leads in 2 of our five dimensions. On price, Beam 1.0 runs about $1.60 per 1M tokens (blended) and is open-weight; Devstral 2 is about $0.88 and open-weight.

Where Beam 1.0 falls short
  • Every published benchmark figure is Reflection's own so far
  • Text only: no image input
Full Beam 1.0 breakdown →
Where Devstral 2 falls short
  • 0% in Artificial Analysis's Terminal-Bench 4.0 run
  • 262K context, smaller than 11.7% of the coding-agent requests we measured
Full Devstral 2 breakdown →

The agent matters as much as the model. Compare AI agents →

Current Claude, GPT, Gemini and open models run on Standard Compute

Standard Compute

Current frontier models Claude Fable 5.1 · GPT-6 Astra

More intelligence
per dollar.

Each step goes to the cheapest model that can do it, on providers we vet for price, uptime and speed, with automatic fallback when one fails.

Compliance
Opt-in rules keep a company, team or key on European or US providers
Privacy
We don’t train on your prompts or store them

Fixed monthly plans from $19 · cancel anytime

7-day fair refund Free trial, no card Billing by Stripe3,500+ agent users

Related comparisons