OpenRouter Pricing Comparison: Four Models and Their Routes

OpenRouter charges the serving provider’s token price for each request and adds 5.5% when credits are bought on Standard, or 8% on Business. Its cheapest GLM-5.3 route, Wafer, costs $0.00210 for 10,000 input and 1,000 output tokens, $0.00221 all-in; the GPT-OSS 120B floor is $0.000475 all-in, just above CoreWeave’s $0.000470 direct. Novita AI’s own routes sit below its direct prices on all three 2026 models.

Updated · 21 sources · Methodology

NeedPickWhy
Cheapest all-in route · GLM-5.3Wafer$0.00210 listed, $0.00221 all-in21
Cheapest all-in route · DeepSeek V4.1 Flash 1017Reka$0.000900 listed, $0.000949 all-in31
Cheapest all-in route · GLM-5.3-FlashRelace$0.000489 listed, $0.000516 all-in41
Cheapest all-in route · GPT-OSS 120BNear AI$0.000450 listed, $0.000475 all-in, tied with Venice51

How OpenRouter pricing works

OpenRouter does not host models. It sends each request to a provider and bills the token price that provider publishes on OpenRouter, then takes its own cut when credits are purchased. A route’s real cost is therefore the route price plus the credit fee, and third-party routes can undercut both first-party APIs and the model labs’ own list prices.

Swipe to see all columns

ChargeRateCharged on1
Standard credits5.5%Each credit purchase, not each request
Business credits8%Each credit purchase on the Business plan
Your own provider key (BYOK)0%, then 5%First $25,000 a month of list-price inference free

Every all-in figure on this page multiplies the route example by 1.055, the Standard fee. On Business credits the same GLM-5.3 route would come to $0.00227.

Cheapest route against buying direct

For each model the table pairs the cheapest OpenRouter route after the 5.5% fee with the cheapest provider list price for the same model. The route wins on the three 2026 models; on GPT-OSS 120B the fee hands the win to CoreWeave.

Swipe to see all columns

MetricCheapest routeRoute exampleAll-in +5.5%Cheapest direct listDirect example
GLM-5.3Wafer2$0.00210$0.002211DeepInfra10$0.01300
DeepSeek V4.1 Flash 1017Reka3$0.000900$0.0009491DeepInfra11$0.00260
GLM-5.3-FlashRelace4$0.000489$0.0005161Seven-way tie8$0.00200
GPT-OSS 120BNear AI5$0.000450$0.0004751CoreWeave17$0.000470
Cheapest all-in cost per model after the 5.5% fee (10,000 input + 1,000 output tokens, USD)
GLM-5.3$0.00221
DeepSeek V4.1 Flash 1017$0.000949
GLM-5.3-Flash$0.000516
GPT-OSS 120B$0.000475

GLM-5.3: OpenRouter routes and direct prices

48 routes price GLM-5.3 on OpenRouter. The route column is the provider’s published OpenRouter price, the direct column that provider’s own list price, and the lowest listed route appears as its own row.

Swipe to see all columns

ProviderRoute in / out per 1MRoute exampleAll-in +5.5%Direct list in / out per 1MDirect example
Novita AI26$0.63 / $1.98$0.00828$0.00874$1.40 / $4.40$0.01840
DeepInfra210$0.5625 / $2.50$0.00813$0.00857$0.90 / $4.00$0.01300
Together AI214$1.40 / $4.40$0.01840$0.01941$1.40 / $4.40$0.01840
Fireworks215$1.40 / $4.40$0.01840$0.01941$1.40 / $4.40$0.01840
Baseten216$1.40 / $4.40$0.01840$0.01941$1.40 / $4.40$0.01840
CoreWeave2$1.20 / $4.00$0.01600$0.01688——
Z.ai218$1.40 / $4.40$0.01840$0.01941$1.40 / $4.40$0.01840
Lowest listed route
wafer/us2		$0.0699 / $1.40$0.00210$0.00221——

DeepInfra lists $0.90/$4.00 for GLM-5.3 and its route shows the current promotion at $0.5625/$2.50. CoreWeave publishes no Forge price for GLM-5.3, so it has a route here and no direct column.

DeepSeek V4.1 Flash 1017: OpenRouter routes and direct prices

35 routes price DeepSeek V4.1 Flash 1017, and the spread is wide: $0.000900 for the 10,000/1,000 example on the cheapest route against $0.00420 at list on Together AI, Fireworks, Baseten and DeepSeek.

Swipe to see all columns

ProviderRoute in / out per 1MRoute exampleAll-in +5.5%Direct list in / out per 1MDirect example
Novita AI37$0.24 / $0.96$0.00336$0.00354$0.30 / $1.20$0.00420
DeepInfra311$0.098 / $0.294$0.00127$0.00134$0.20 / $0.60$0.00260
Together AI314$0.30 / $1.20$0.00420$0.00443$0.30 / $1.20$0.00420
Fireworks315$0.30 / $1.20$0.00420$0.00443$0.30 / $1.20$0.00420
Baseten316$0.30 / $1.20$0.00420$0.00443$0.30 / $1.20$0.00420
DeepSeek319$0.30 / $1.20$0.00420$0.00443$0.30 / $1.20$0.00420
Lowest listed route
reka3		$0.04 / $0.50$0.000900$0.000949——

DeepInfra lists $0.20/$0.60 for DeepSeek V4.1 Flash 1017 and its route carries the current promotion at $0.098/$0.294; a Flex route at $0.0784/$0.2352 is cheaper still. DeepSeek bills $0.15/$0.60 off-peak against $0.30/$1.20 at peak.

GLM-5.3-Flash: OpenRouter routes and direct prices

38 routes price GLM-5.3-Flash. Six of the seven providers in the table route at the same $0.15/$0.50, so the gap comes from the discount routes and from Novita AI’s own route at $0.084/$0.28.

Swipe to see all columns

ProviderRoute in / out per 1MRoute exampleAll-in +5.5%Direct list in / out per 1MDirect example
Novita AI48$0.084 / $0.28$0.00112$0.00118$0.15 / $0.50$0.00200
DeepInfra412$0.15 / $0.50$0.00200$0.00211$0.15 / $0.50$0.00200
Together AI414$0.15 / $0.50$0.00200$0.00211$0.15 / $0.50$0.00200
Fireworks415$0.15 / $0.50$0.00200$0.00211$0.15 / $0.50$0.00200
Baseten416$0.15 / $0.50$0.00200$0.00211$0.15 / $0.50$0.00200
CoreWeave417$0.15 / $0.50$0.00200$0.00211$0.15 / $0.50$0.00200
Z.ai418$0.15 / $0.50$0.00200$0.00211$0.15 / $0.50$0.00200
Lowest listed route
relace4		$0.0413 / $0.0758$0.000489$0.000516——

DeepInfra’s GLM-5.3-Flash route of $0.15/$0.50 matches its current price; AA’s older $0.075/$0.25 listing is not used. CoreWeave caches at $0.02 on OpenRouter against $0.05 on its own page.

GPT-OSS 120B: OpenRouter routes and direct prices

24 routes price GPT-OSS 120B. Its direct prices are already low, so the 5.5% fee is almost the whole story: the cheapest route and the cheapest direct provider land about 1% apart.

Swipe to see all columns

ProviderRoute in / out per 1MRoute exampleAll-in +5.5%Direct list in / out per 1MDirect example
Novita AI59$0.05 / $0.25$0.000750$0.000791$0.05 / $0.25$0.000750
DeepInfra513$0.037 / $0.17$0.000540$0.000570$0.037 / $0.17$0.000540
Together AI514$0.15 / $0.60$0.00210$0.00222$0.15 / $0.60$0.00210
Fireworks5—————
Baseten5$0.10 / $0.50$0.00150$0.00158——
CoreWeave517$0.03 / $0.17$0.000470$0.000496$0.03 / $0.17$0.000470
Groq520$0.15 / $0.60$0.00210$0.00222$0.15 / $0.60$0.00210
Cerebras521$0.35 / $0.75$0.00425$0.00448$0.35 / $0.75$0.00425
Lowest listed route
near-ai/fp45		$0.03 / $0.15$0.000450$0.000475——

Fireworks sells GPT-OSS 120B at $0.15/$0.60 but has no route in this snapshot. CoreWeave’s route and direct price are identical at $0.03/$0.17.

Where each provider’s routes land

Cards follow the GLM-5.3 route example upward, standard routes only, with OpenRouter first because it is the subject of this page. All-in figures are the route example for 10,000 input and 1,000 output tokens after the 5.5% Standard fee.

  1. OpenRouter

    The router: 145 routes across the four models, each billed at the serving provider’s published token price.12345

    GLM-5.3 all-in
    $0.00221
    DeepSeek V4.1 Flash 1017 all-in
    $0.000949
    GLM-5.3-Flash all-in
    $0.000516

    Strengths

    Its cheapest route is also the cheapest all-in option on all three 2026 models: $0.00221 for GLM-5.3, $0.000949 for DeepSeek V4.1 Flash 1017 and $0.000516 for GLM-5.3-Flash.12345

    Limitations

    The cheapest route on each model is a small host with no published capacity data, and the 8% Business fee lifts the GPT-OSS 120B floor to $0.000486, above CoreWeave’s $0.000470 direct.12345

    12345

  2. DeepInfra

    The cheapest named route on GLM-5.3: $0.00857 all-in against a $0.01300 direct list.23410111213

    GLM-5.3 all-in
    $0.00857
    DeepSeek V4.1 Flash 1017 all-in
    $0.00134
    GLM-5.3-Flash all-in
    $0.00211

    Strengths

    Its DeepSeek V4.1 Flash 1017 route is the cheapest named route on that model at $0.00134 all-in, and its direct list of $0.90/$4.00 is the lowest standard GLM-5.3 price among direct sellers.23410111213

    Limitations

    Both low figures are current promotions, not standing prices: GLM-5.3 runs $0.5625/$2.50 against a $0.90/$4.00 list, and DeepSeek V4.1 Flash 1017 runs $0.098/$0.294 against $0.20/$0.60.23410111213

    23410111213

  3. Novita AI

    Its OpenRouter routes cost less than its own list: $0.00874 all-in on GLM-5.3 against $0.01840 direct.2346789

    GLM-5.3 all-in
    $0.00874
    DeepSeek V4.1 Flash 1017 all-in
    $0.00354
    GLM-5.3-Flash all-in
    $0.00118

    Strengths

    Its routes undercut its direct prices on all three 2026 models: GLM-5.3 at $0.63/$1.98 against $1.40/$4.40, DeepSeek V4.1 Flash 1017 at $0.24/$0.96 against $0.30/$1.20, and GLM-5.3-Flash at $0.084/$0.28 against $0.15/$0.50.2346789

    Limitations

    Buying direct costs more than its own route: $0.01840 per GLM-5.3 example against $0.00828 through OpenRouter, and $0.00420 against $0.00336 on DeepSeek V4.1 Flash 1017. Its route is still not the cheapest on any model: $0.00210 on GLM-5.3, $0.000900 on DeepSeek V4.1 Flash 1017 and $0.000489 on GLM-5.3-Flash cost less.2346789

    2346789

  4. CoreWeave

    A route with no matching public price: $0.01688 all-in on GLM-5.3, but CoreWeave publishes no Forge rate for that model.2417

    GLM-5.3 all-in
    $0.01688
    DeepSeek V4.1 Flash 1017 all-in
    —
    GLM-5.3-Flash all-in
    $0.00211

    Strengths

    On GPT-OSS 120B its direct $0.03/$0.17 is the cheapest all-in price on this page at $0.000470, just under OpenRouter’s $0.000475 all-in floor.2417

    Limitations

    GLM-5.3 is available only through its OpenRouter route at $1.20/$4.00, and it has no DeepSeek V4.1 Flash 1017 route at all.2417

    2417

  5. Baseten

    Full list on every model, with the lowest cached-input rate among the standard GLM-5.3 routes at $0.14.23416

    GLM-5.3 all-in
    $0.01941
    DeepSeek V4.1 Flash 1017 all-in
    $0.00443
    GLM-5.3-Flash all-in
    $0.00211

    Strengths

    Its $0.14 cached rate on GLM-5.3 beats Novita AI, Together AI, Fireworks and Z.ai at $0.26, which matters for long prompts with repeated prefixes.23416

    Limitations

    No route discount at all: $1.40/$4.40 on GLM-5.3 and $0.30/$1.20 on DeepSeek V4.1 Flash 1017, the same as the labs charge.23416

    23416

  6. Together AI

    List-price parity: $0.01941 all-in on GLM-5.3, the same figure as Baseten and Z.ai.23414

    GLM-5.3 all-in
    $0.01941
    DeepSeek V4.1 Flash 1017 all-in
    $0.00443
    GLM-5.3-Flash all-in
    $0.00211

    Strengths

    Its route and direct prices agree at $1.40/$4.40 for GLM-5.3 and $0.30/$1.20 for DeepSeek V4.1 Flash 1017, so the 5.5% fee is the only difference between routing and buying direct.23414

    Limitations

    None of its three current-model routes is a discount: the GLM-5.3 route costs $0.01840 per example against $0.00813 on the DeepInfra promotion and $0.00210 on the cheapest route.23414

    23414

  7. Fireworks

    Two tiers, two prices: its standard GLM-5.3 route is $0.01941 all-in and its spot route $0.01359.23415

    GLM-5.3 all-in
    $0.01941
    DeepSeek V4.1 Flash 1017 all-in
    $0.00443
    GLM-5.3-Flash all-in
    $0.00211

    Strengths

    The spot route takes GLM-5.3 to $0.98/$3.08, 30% below the $1.40/$4.40 that Together AI, Baseten and Z.ai list on the same model.23415

    Limitations

    Spot is a separate tier, not the published standard rate, and DeepSeek V4.1 Flash 1017 splits the same way: $0.21/$0.84 spot against $0.30/$1.20 standard.23415

    23415

  8. Z.ai

    The model lab on its own route: $0.01941 all-in on GLM-5.3, in a four-way tie at list price.2418

    GLM-5.3 all-in
    $0.01941
    DeepSeek V4.1 Flash 1017 all-in
    —
    GLM-5.3-Flash all-in
    $0.00211

    Strengths

    First-party GLM-5.3 and GLM-5.3-Flash at list prices, with a $0.26 cached rate on the flagship equal to Novita AI and Together AI.2418

    Limitations

    Its GLM-5.3 route costs $0.01840 per example against $0.00210 on the cheapest Wafer route, so the same request is 8.8× dearer at list.2418

    2418

Methodology and sources

OpenRouter route prices come from each model’s published endpoint list, read on 2026-10-26; the all-in column multiplies the 10,000-input/1,000-output example by 1.055, the Standard credit-purchase fee from OpenRouter’s pricing page. Direct columns are providers’ own standard serverless list prices, checked on their own pages the same day; where a provider runs a current promotion the list price is shown and the promotion is named in the card.

Costs are estimates at listed rates for uncached synchronous requests, not measured invoices. Tax, retries, discounts and negotiated capacity are excluded, and cached-input rates are left out of the example because they depend on prefix reuse. Novita AI publishes this page from public sources, without a private workload benchmark.

Rows inside each model table keep the same provider order on every model, starting with Novita AI, and are not a price ranking; the lowest listed route of any host appears on its own row. Cards are ordered by the standard-route GLM-5.3 all-in example, with OpenRouter first as the subject of the page; the Fireworks spot tier is noted in its card but does not set the order. Route prices can change without notice because they belong to the serving provider.1

  1. OpenRouter · Plans and credit-purchase fees
  2. OpenRouter · GLM-5.3 routes
  3. OpenRouter · DeepSeek V4.1 Flash 1017 routes
  4. OpenRouter · GLM-5.3-Flash routes
  5. OpenRouter · GPT-OSS 120B routes
  6. Novita AI · GLM-5.3
  7. Novita AI · DeepSeek V4.1 Flash 1017
  8. Novita AI · GLM-5.3-Flash
  9. Novita AI · GPT-OSS 120B
  10. DeepInfra · GLM-5.3
  11. DeepInfra · DeepSeek V4.1 Flash 1017
  12. DeepInfra · GLM-5.3-Flash
  13. DeepInfra · GPT-OSS 120B
  14. Together AI · pricing
  15. Fireworks · serverless pricing
  16. Baseten · pricing
  17. CoreWeave · Forge token prices
  18. Z.ai · pricing
  19. DeepSeek · API pricing
  20. Groq · GPT-OSS 120B
  21. Cerebras · pricing

Build with Novita Model APIs

Create an API key and connect your application.

Start BuildingExplore Models