Cost per coding-agent request

Is Devstral 2 or GLM 4.6 cheaper for coding agents?

For coding agents, GLM 4.6 costs 1.7x as much as Devstral 2 per request ($0.015 vs $0.0089 on our measured traffic), because GLM 4.6 lists at $0.50 input, $0.10 cached input and $2 output per million tokens against $0.40, $0.04 and $2 for Devstral 2.

List prices: OpenRouter listed price, checked 2026-10-13. Request costs: Standard Compute gateway data, 79,301 coding-agent requests from 75 users, Sep 13 to Oct 12, 2026. How we compute this

Run current Claude, GPT, Gemini and open models in your agent

Or let the router send each step to the cheapest model that can do it, on providers we vet for price, uptime and speed. Plans from $19 a month.

See plans

Cost per coding-agent request, side by side

Cost at list price	Devstral 2	GLM 4.6
Prompt under 100K tokens50% of requests, 87% of input cached	$0.0062	$0.0100
Prompt of 100K to 272K39% of requests, 88% of input cached	$0.014	$0.023
Prompt over 272K tokens11% of requests, 86% of input cached	n/a	n/a
Average request	$0.0089	$0.015
100 requests	$0.89	$1.47
Median session16 requests; a session ends after 10 idle minutes	$0.11	$0.17
A month of regular usemedian of 20 users active on 5+ days	$10.02	$15.56

Each logged request priced at the listed rates above for its own token counts, including cache reads and any higher rate for long prompts. Average output: 698 tokens per request. Both models are priced on the requests up to 205K tokens that fit in both context windows; band shares above describe all requests.

Benchmarks, speed and latency, side by side

Artificial Analysis	Devstral 2	GLM 4.6
Intelligence Index10 evaluations, 0 to 100	9	19 (estimate)
Terminal-Bench 4.0agentic coding in a terminal	0.0%	n/a
Output speedmedian	n/a	36 tokens/s
Time to first answer tokenincludes thinking time	n/a	59 s

Source: Artificial Analysis (artificialanalysis.ai). Measured variants: Devstral 2; GLM-4.6 (Reasoning). Fetched October 13, 2026. Artificial Analysis runs Terminal-Bench 4.0 in its own setup, so its scores differ from the tbench.ai leaderboard.

What to watch for

GLM 4.6's 205K window cannot take 20% of coding-agent requests, so the comparison uses only requests both can take. 87% of agent input is read from cache, and GLM 4.6 charges $0.10 per million for it against $0.04. The price is OpenRouter's listed rate on October 13, 2026, which can be a third-party host's; Z.AI's own API may charge differently.

FAQ

Is Devstral 2 cheaper than GLM 4.6 for coding agents?

Yes, at the listed prices. 100 average coding-agent requests cost $0.89 on Devstral 2 and $1.47 on GLM 4.6. A median 16-request session costs $0.11 vs $0.17. Measured on 79,301 real coding-agent requests (Sep 13 to Oct 12, 2026), counting only the requests up to 205K tokens that both can take.

How much does a month of GLM 4.6 vs Devstral 2 cost?

For the median regular user in our data (1,478 requests in 30 days), about $15.56 on GLM 4.6 and $10.02 on Devstral 2 at list price. Heavier users pay proportionally more.

Devstral 2 vs GLM 4.6: which is better for agents?

GLM 4.6 scores higher on output quality (4.8/10 vs 3.3/10), while Devstral 2 is the cheaper option. Devstral 2 is best for coding agents that need a Mistral open-weight model; GLM 4.6 is best for short-context coding tasks on a tight budget. Live community votes on this page show what people actually prefer.

Can I switch between Devstral 2 and GLM 4.6 in the same agent?

Yes. Model-agnostic agents (Cline, Aider, OpenClaw, Hermes Agent, Kilo Code, OpenCode) let you swap models per task. Many people route hard tasks to the stronger model and routine work to the cheaper one.

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

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

Get your API key See how it works

Fixed monthly plans from $19 · cancel anytime

7-day fair refund Free trial, no card Billing by Stripe4,100+ agent users